Hermes-Google is an MCP server that gives an AI assistant (e.g., Claude/Hermes) scoped access to Gmail, Google Calendar, and Google Drive through a dedicated Hermes Google account, keeping it isolated from your personal Google account. All write operations require user confirmation.
Authentication
auth_status — Check whether stored OAuth credentials are valid, expired, and what scopes are granted
Gmail
mail_list_pending — List unread forwarded emails (newest first, up to 100)
mail_search — Search emails using Gmail search syntax (e.g., from:alice, subject:invoice)
mail_get — Fetch a single email by ID, including full content, attachments, and threading info
mail_send_draft — Send a draft email from Hermes's account, restricted to the user's own inbox only (external recipients blocked)
mail_mark_read — Mark a message as read
mail_archive — Archive a message (removes INBOX label)
Google Calendar
cal_list_calendars — List all calendars visible to Hermes (own + shared)
cal_list_events — List events in a calendar within a time range
cal_create_event — Create a calendar event with title, time, attendees, and description
cal_update_event — Patch specific fields of an existing event
cal_delete_event — Delete a calendar event
Google Drive
drive_search — Search Drive files by name, with optional MIME type filter
drive_list — List the contents of a Drive folder by folder ID
drive_get — Download a Drive file to local cache (~/.cache/hermes-google/drive/)
drive_upload — Upload a local file to Drive (with optional target folder)
drive_update — Replace an existing Drive file's contents with a local file
drive_move — Move a Drive file to a different parent folder
drive_delete — Delete a Drive file
Allows reading and managing emails, including listing pending emails, searching, getting messages, sending drafts, marking read, and archiving.
Allows interacting with Google Calendar, including listing calendars, listing events, and creating, updating, and deleting events.
Allows searching, listing, uploading, updating, moving, and deleting files in Google Drive.
Click on "Install Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type @ followed by the MCP server name and your instructions, e.g., "@hermes-google check my unread emails"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
MCP server giving Hermes (Claude Code running as a personal assistant) scoped access to Gmail, Google Calendar, and Google Drive through a dedicated Hermes Google account — without granting access to your personal Google account.
See docs/superpowers/specs/2026-04-23-hermes-google-design.md
for the full design.
Add hermes-google to your project's .mcp.json:
{
"mcpServers": {
"hermes-google": {
"type": "stdio",
"command": "uvx",
"args": ["hermes-google-mcp@latest"]
}
}
}Related MCP server: gmail-drive-mcp-server
Sign up for a plain Gmail account (e.g., hermes-yourname@gmail.com)
In Google Cloud Console: create a project, enable the Gmail, Calendar, and Drive APIs
Create an OAuth 2.0 Client ID (type: Desktop application) and download the client secret JSON
Save it as ~/.config/hermes-google/oauth_client.json
Run the setup script:
git clone https://github.com/jimmy-larsson/hermes-google.git
cd hermes-google
./scripts/setup.shThe setup script creates the config, runs the OAuth flow (saving the token to
~/.config/hermes-google/token.json), and prints the remaining manual steps.
The OAuth flow opens a browser for consent — it must run on your host machine, not inside a headless container.
If your config and credentials live inside a container (e.g. Docker-mounted
~/.config/hermes-google/), you need to:
Copy config.toml and oauth_client.json to the host filesystem at
~/.config/hermes-google/
Run the OAuth flow from the host:
uvx hermes-google-mcp@latest auth loginThe resulting token.json is saved on the host — it will be picked up by
the container through the volume mount
Create filters in your personal Gmail to route emails to the Hermes account:
For each sender you want Hermes to handle: create a filter that both
labels the email (e.g. hermes-review) and forwards it to
hermes-yourname@gmail.com — in a single filter
For replies from Hermes: from:hermes-yourname@gmail.com → apply label
hermes
Filters don't chain. Gmail evaluates all filters in a single pass against the original message properties. A label applied by filter A will not trigger filter B that matches on that label. You must combine label + forward into one filter per sender/criteria.
"Apply to existing" skips forwarding. When you click "Also apply filter to matching conversations", Gmail only runs local actions (label, archive, star). Forwarding only fires on new incoming messages.
Calendar: Share your calendar with the Hermes account at "Make changes to
events" permission level. Set [user].calendar_id in config.toml.
Drive: Share specific files/folders with the Hermes account. Optionally
set [drive].default_parent_folder_id in config.toml for a default upload
folder.
Once installed, the following tools are available to Hermes in every session:
mail_list_pending, mail_search, mail_get, mail_send_draft,
mail_mark_read, mail_archive
cal_list_calendars, cal_list_events, cal_create_event,
cal_update_event, cal_delete_event
drive_search, drive_list, drive_get, drive_upload, drive_update,
drive_move, drive_delete
auth_status
All write operations require user confirmation. mail_send_draft is
structurally restricted to your own email; it cannot send to external
recipients.
Same operations via shell:
hermes-google auth status
hermes-google mail list --limit 10
hermes-google mail get <message_id>
hermes-google cal list --start 2026-04-24T00:00:00+09:00 --end 2026-04-25T00:00:00+09:00
hermes-google drive search "Q1 report"Any one of these fully cuts an integration surface:
Delete the forwarding filter in your personal Gmail
Unshare your calendar with the Hermes account
Unshare a Drive file or folder
hermes-google auth revoke — removes the refresh token locally
claude mcp remove hermes-google — Hermes loses the tools; Google data
untouched
Delete the Hermes Google account entirely
conda activate hermes-google
pytest
ruff check src tests# 1. Bump version in pyproject.toml, commit and push
git commit -am "chore: bump version to X.Y.Z"
git push
# 2. Tag locally and push — triggers the CI pipeline
git tag X.Y.Z
git push origin X.Y.Z
# 3. After CI passes, create the GitHub release
gh release create X.Y.Z --verify-tag --generate-notes# User scope — available in all your projects
claude mcp add -s user hermes-google -- uvx hermes-google-mcp
# Local scope — private to you in this project only
claude mcp add hermes-google -- uvx hermes-google-mcpReport whether stored credentials are valid and loaded.
Returns dict with keys: valid (bool), expired (bool), scopes (list of
granted OAuth scope URLs), and — only on failure — error (string describing
the auth or config problem).
| Name | Required | Description | Default |
|---|---|---|---|
No parameters | |||
| Name | Required | Description |
|---|---|---|
No output parameters | ||
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
The description goes beyond annotations (readOnlyHint, destructiveHint) by detailing the actual returned keys (valid, expired, scopes, error) and conditions (error only on failure). No contradictions.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is very concise (two sentences) and well-structured, with key information front-loaded and return keys listed clearly.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Given no parameters, good annotations, and a detailed output description, the tool definition is fully complete. The output schema is implied but covered by the description.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
No parameters exist, but the description adds significant value by explaining the return value structure in detail. This compensates fully for the lack of parameters.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description explicitly states the tool's purpose: 'Report whether stored credentials are valid and loaded.' It then details the return dictionary with keys, which clearly distinguishes it from sibling tools that perform CRUD operations on calendars, drives, and mail.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
The description implies usage for checking auth status before other operations but does not give explicit when-to-use or when-not-to-use guidance. However, the context (sibling tools are resource-specific) makes the use case clear.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Create a calendar event. Requires user confirmation before calling.
calendar accepts 'user', 'hermes', or a full Google Calendar ID.
start and end must be RFC 3339 datetimes with timezone offset.
Returns dict with key: id (created event ID).
| Name | Required | Description | Default |
|---|---|---|---|
| end | Yes | ||
| start | Yes | ||
| title | Yes | ||
| calendar | Yes | ||
| attendees | No | ||
| description | No |
| Name | Required | Description |
|---|---|---|
No output parameters | ||
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
The description adds important behavioral context beyond annotations by noting the need for user confirmation and describing the return value (dict with 'id'). Annotations do not contradict; destructiveHint=false aligns with creation.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is very concise, consisting of 5 short lines. It front-loads the purpose and important note, with every sentence adding value.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
The description covers the key parameters and return value reasonably well. Minor gaps remain in explaining `title` and optional parameters, but given the tool's simplicity and existence of an output schema, it is fairly complete.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
With 0% schema coverage, the description compensates by explaining `calendar` acceptable values and `start`/`end` format. However, it does not describe `title`, `attendees`, or `description`, leaving some gaps.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states 'Create a calendar event.' This is a specific verb and resource, and it distinguishes this tool from siblings like cal_update_event and cal_delete_event.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
The description explicitly says 'Requires user confirmation before calling.' and provides formatting instructions for `start` and `end`. It does not explicitly mention when not to use or alternatives, but the context is clear.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Delete a calendar event. Requires user confirmation before calling.
calendar accepts 'user', 'hermes', or a full Google Calendar ID.
Returns dict with key: id (deleted event ID).
| Name | Required | Description | Default |
|---|---|---|---|
| calendar | Yes | ||
| event_id | Yes |
| Name | Required | Description |
|---|---|---|
No output parameters | ||
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
Annotations already indicate destructiveHint: true. The description adds the important behavioral trait of requiring user confirmation, which goes beyond the annotation. It also describes the return value, but does not disclose error handling or side effects.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is concise with three sentences covering purpose, parameter guidance, and return value. No redundant information, well-structured.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Given the tool's complexity (2 params, output schema), the description is mostly complete. It covers the return value and parameter specifics for calendar, but omits handling of non-existent events or authorization requirements. Still, it is adequate for most use cases.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
With 0% schema description coverage, the description must compensate. It partially explains the `calendar` parameter (accepts 'user', 'hermes', or full ID), but leaves `event_id` unexplained. This is insufficient for a low-coverage scenario.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states 'Delete a calendar event' with a specific verb and resource. It distinguishes from sibling tools like cal_create_event and cal_update_event by focusing on deletion.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
The description explicitly mentions 'Requires user confirmation before calling,' providing a clear usage guideline. It also explains the `calendar` parameter values, but lacks guidance on prerequisites like event existence or permissions.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
List all calendars visible to Hermes (own calendar + any shared with Hermes).
Returns a list of dicts, each with keys: id (calendar ID), summary
(display name), access_role (e.g., 'owner', 'writer', 'reader').
| Name | Required | Description | Default |
|---|---|---|---|
No parameters | |||
| Name | Required | Description |
|---|---|---|
| result | Yes |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
Annotations already mark it as read-only and non-destructive. The description adds the exact return structure (list of dicts with keys: id, summary, access_role), which goes beyond annotations to clarify output behavior.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
Two sentences, front-loaded with the core action, no extraneous words. Every sentence adds value.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
With zero parameters, full annotations, and an output schema (implied by context), the description is complete. It describes the return format and scope, leaving no gaps.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
No parameters exist, and schema coverage is 100%. The description does not need to explain parameters; it adds no extra param info, which is appropriate given the baseline of 4 for zero parameters.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states 'List all calendars visible to Hermes', specifying the verb (list) and resource (calendars) with scope (own + shared). This distinguishes it from sibling tools that handle events, drive, mail, or auth.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
The description provides no explicit guidance on when to use this tool versus alternatives or when not to use it. The use case is implied by the purpose, but lacking exclusions or context, it gets a moderate score.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
List events in a calendar within a time range, sorted chronologically.
calendar accepts 'user' (your personal calendar), 'hermes' (Hermes's
primary calendar), or a full Google Calendar ID.
start and end must be RFC 3339 datetimes with timezone offset
(e.g., '2026-04-24T00:00:00+08:00'). Bare dates are rejected by the API.
Returns a list of dicts, each with keys: id, title, start, end,
attendees (list of email strings).
| Name | Required | Description | Default |
|---|---|---|---|
| end | Yes | ||
| start | Yes | ||
| calendar | Yes |
| Name | Required | Description |
|---|---|---|
| result | Yes |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
Annotations already indicate read-only and non-destructive behavior. The description adds value by specifying the return format (list of dicts with keys) and input constraints (RFC 3339 datetimes, rejection of bare dates), which are beyond the annotations.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is front-loaded with the purpose, then explains each parameter in a structured manner. No redundant sentences; every sentence provides essential information efficiently.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Given the moderate complexity of the tool (3 required parameters, no nested objects) and the presence of an output schema (return format explained in description), the description covers all necessary aspects: input format, valid values, and output structure.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
Schema description coverage is 0%, but the description fully compensates by detailing acceptable values for 'calendar' (user, hermes, full ID) and exact format for 'start' and 'end' (RFC 3339 with timezone offset, rejecting bare dates). This adds significant meaning beyond the schema.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states the tool lists events in a calendar within a time range, sorted chronologically. This specific verb+resource combination distinguishes it from siblings like cal_create_event or cal_delete_event.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
The description provides clear context for using the tool (time range, accepted calendar IDs, required datetime format) but does not explicitly contrast with alternatives or state when not to use it. The purpose is clear from the name and context.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Patch specific fields of a calendar event. Requires user confirmation before calling.
calendar accepts 'user', 'hermes', or a full Google Calendar ID.
fields is a dict of Google Calendar event fields to update (e.g.,
{"summary": "New title", "location": "Room 3"}).
Returns dict with key: id (updated event ID).
| Name | Required | Description | Default |
|---|---|---|---|
| fields | Yes | ||
| calendar | Yes | ||
| event_id | Yes |
| Name | Required | Description |
|---|---|---|
No output parameters | ||
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
Annotations already mark the tool as idempotent and non-destructive. The description adds the critical behavioral trait 'Requires user confirmation before calling', which is not covered by annotations. It also explains the return structure. This is good supplementary context.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is concise, with a clear primary sentence followed by brief parameter explanations. No redundant information. Every sentence serves a purpose.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Given the presence of an output schema (mentioned), the description adequately covers the return format (dict with 'id'). It addresses key usage constraints (user confirmation, calendar ID format) and provides an example for the fields parameter. No critical gaps.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
Schema description coverage is 0%. The description provides meaning for 'calendar' (accepted values: 'user', 'hermes', or full Google Calendar ID) and 'fields' (dict of event fields with example). 'event_id' is not described but it is a straightforward required parameter. The description adds substantial value beyond the bare schema.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description states 'Patch specific fields of a calendar event', which clearly specifies the verb (patch) and resource (calendar event). It distinguishes from siblings like cal_create_event (create) and cal_delete_event (delete) by focusing on updating existing events.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
The description implies the tool should be used for partial updates, though it does not explicitly state when to use versus alternatives. It mentions a prerequisite (user confirmation) but no explicit when-not or alternative guidance. Given no sibling for full update, it is sufficiently clear.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Delete a Drive file. Requires user to have used an explicit 'delete' verb.
Returns dict with key: id (deleted Drive file ID).
| Name | Required | Description | Default |
|---|---|---|---|
| file_id | Yes |
| Name | Required | Description |
|---|---|---|
No output parameters | ||
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
Annotations already indicate destructiveHint=true. The description adds that the user needs an explicit 'delete' verb and that the return value is a dict with 'id'. This provides context beyond annotations.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is concise: two sentences conveying purpose, requirement, and return value with no redundant information.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
For a simple delete tool with one parameter and an output schema, the description is mostly adequate but lacks clarification on whether deletion is permanent or trash, and what 'explicit delete verb' entails in terms of permissions.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
The input schema has 0% description coverage for the only parameter 'file_id', and the description does not elaborate on its meaning or format. This is a significant gap.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states 'Delete a Drive file', which is a specific verb and resource. This distinguishes it from sibling tools like drive_get, drive_list, and drive_update.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
It mentions a requirement: 'Requires user to have used an explicit delete verb', providing a precondition. However, it does not explicitly state when to use this tool versus alternatives or when not to use it.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Download a Drive file to the local cache directory.
The file is saved to ~/.cache/hermes-google/drive//. Use Claude Code's Read tool on the returned path to view the contents.
Returns dict with key: path (absolute local file path).
| Name | Required | Description | Default |
|---|---|---|---|
| file_id | Yes |
| Name | Required | Description |
|---|---|---|
No output parameters | ||
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
The description aligns with annotations (readOnlyHint=true, destructiveHint=false) by describing a read operation. It adds details about file caching location and return format, which are beyond what annotations provide, enhancing transparency.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is concise with three sentences, each serving a clear purpose: stating the action, explaining the cache path, and describing the return value. It is front-loaded and efficient.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Given the simple input (one parameter), strong annotations, and presence of an output schema, the description is complete. It explains the return dict with the 'path' key, fulfilling the expected information.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
With 0% schema description coverage, the description should explain the parameter semantics. 'file_id' is self-explanatory as a file identifier, but the description does not elaborate on its format or constraints, providing marginal added value.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states the verb 'Download' and the resource 'a Drive file', specifying the action and target. It distinguishes from siblings like drive_list or drive_search by focusing on downloading a single file.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
The description indicates the tool saves to a local cache and advises using the Read tool to view contents, providing practical usage guidance. However, it does not explicitly state when to use or avoid this tool versus alternatives, leaving some ambiguity.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
List children of a Drive folder by folder ID.
Returns a list of dicts, each with keys: id, name, mime_type.
| Name | Required | Description | Default |
|---|---|---|---|
| limit | No | Max results (1-100) | |
| folder_id | Yes |
| Name | Required | Description |
|---|---|---|
| result | Yes |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
Annotations already indicate read-only and non-destructive behavior. The description adds the return format (list of dicts with specific keys), providing extra context beyond annotations.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
Two sentences, front-loaded with the core action, no unnecessary information. Every sentence serves a purpose.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Given the presence of an output schema, the description does not need to detail return values fully. It mentions the structure but does not cover pagination or error handling, which are minor gaps.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
Schema description coverage is 50% (only 'limit' described). The description adds meaning for the required 'folder_id' parameter, but overall parameter information is minimal and does not significantly enhance the schema.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states the action ('List children') and the resource ('Drive folder by folder ID'), and distinguishes from siblings like drive_search and drive_get.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
No explicit guidance on when to use this tool versus alternatives like drive_search for searching files or drive_get for a single file. The description implies its use for listing folder contents but does not clarify when not to use it.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Move a Drive file into a different parent folder. Requires user confirmation before calling.
Returns dict with key: id (moved Drive file ID).
| Name | Required | Description | Default |
|---|---|---|---|
| file_id | Yes | ||
| parent_folder_id | Yes |
| Name | Required | Description |
|---|---|---|
No output parameters | ||
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
Annotations already indicate non-destructive and idempotent behavior. The description adds context about user confirmation and return value ('Returns dict with key: id'), which provides additional behavioral insight beyond annotations.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is two sentences, no filler. It is front-loaded with the action and key behavioral note. Slightly more detail on parameters would be welcome, but it remains efficient.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Given the tool has two parameters and an output schema, the description covers the action and return value partially. However, parameter semantics are lacking, and the output schema is not fully described. It is adequate but not thorough.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
Schema coverage is 0%, so the description should compensate, but it only mentions 'Drive file' and 'parent folder' without detailing parameter meaning (e.g., file_id format, parent_folder_id constraints). The description adds minimal semantic value for parameters.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description states 'Move a Drive file into a different parent folder' with a specific verb and resource. It clearly distinguishes from sibling tools like drive_delete or drive_list.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
The description mentions 'Requires user confirmation before calling,' which provides a usage condition. However, it does not give guidance on when not to use the tool or suggest alternative tools for similar tasks.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Search Drive files visible to Hermes by filename.
Optionally filter by MIME type (e.g., 'application/pdf', 'application/vnd.google-apps.spreadsheet').
Returns a list of dicts, each with keys: id, name, mime_type.
| Name | Required | Description | Default |
|---|---|---|---|
| limit | No | Max results (1-100) | |
| query | Yes | ||
| mime_type | No |
| Name | Required | Description |
|---|---|---|
| result | Yes |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
Annotations already declare readOnlyHint=true and destructiveHint=false, so the description need not restate safety. It adds value by specifying the search criteria (by filename) and the return format (list of dicts with id, name, mime_type). However, it does not disclose search behavior details like case sensitivity, partial matching, or pagination, which would be useful.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is two sentences plus a return type note, with no wasted words. It is front-loaded and includes examples, making it easy to parse quickly.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Given the existence of an output schema and annotations, the description covers the basics but leaves ambiguity about search semantics (e.g., exact match vs substring, case sensitivity) and the meaning of 'visible to Hermes'. For a search tool, more detail on behavior would enhance completeness.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
With only 33% schema description coverage (only limit has a description), the description compensates by clarifying that the query parameter is for filename search, and gives examples for mime_type. The limit parameter is already described in the schema, so no extra credit. The description adds meaningful context beyond the raw schema.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states the tool searches Drive files by filename, with optional MIME type filter. This distinguishes it from sibling tools like drive_list (which likely lists all files) and drive_get (retrieves a specific file). The verb 'search' and resource 'Drive files' are specific, and the return structure is provided.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
The description implies usage for finding files by name, but does not explicitly state when to use this tool over alternatives like drive_list or drive_get. No exclusions or when-not-to-use guidance is provided, leaving the agent to infer from the tool name and description.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Replace a Drive file's contents with a local file. Requires user confirmation before calling.
Returns dict with key: id (updated Drive file ID).
| Name | Required | Description | Default |
|---|---|---|---|
| file_id | Yes | ||
| local_path | Yes |
| Name | Required | Description |
|---|---|---|
No output parameters | ||
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
The description adds value beyond annotations by stating the tool 'Requires user confirmation before calling,' which is a critical behavioral caveat not captured in annotations. Annotations declare destructiveHint: false and idempotentHint: true, which are consistent with the described replacement action, and the return format is clearly specified.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is extremely concise—two sentences, no fluff. The first sentence captures the core action and a critical behavioral condition; the second describes the return value. Every sentence earns its place.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Given the tool's simplicity (two parameters, implied output schema, and annotations), the description is largely complete. It covers the action, a key behavioral requirement, and the return format. It does not address error cases or permissions, but these are partly covered by annotations and the tool's straightforward nature.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
With 0% schema description coverage, the description must compensate for parameter meaning. While the action implies that file_id is the target Drive file and local_path is the source file, no explicit parameter descriptions are provided. The names are somewhat self-explanatory, but the description does not fully compensate for the lack of schema documentation.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description states the specific action ('Replace a Drive file's contents with a local file'), clearly identifying the verb and resource. The sibling tools include drive_upload and drive_move, so this description differentiates it as an update operation on existing file content.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
The description does not provide guidance on when to use this tool vs. alternatives like drive_upload or drive_search. It lacks explicit when-to-use or when-not-to-use context, leaving the agent to infer usage from the purpose alone.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Upload a local file to Drive. Requires user confirmation before calling.
If folder_id is omitted, uploads to the configured default parent folder.
Returns dict with key: id (created Drive file ID).
| Name | Required | Description | Default |
|---|---|---|---|
| name | Yes | ||
| folder_id | No | ||
| local_path | Yes |
| Name | Required | Description |
|---|---|---|
No output parameters | ||
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
The description adds behavioral context such as requiring user confirmation and the default folder behavior when folder_id is omitted. While annotations (destructiveHint=false) suggest it is not destructive, the description does not contradict them and provides additional useful information.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is concise with three focused sentences: purpose, condition, and return value. No unnecessary words; each sentence adds value.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
For a simple upload tool, the description covers key aspects: purpose, user confirmation, default folder, and return value. It lacks details on local_path and name semantics but is sufficient for typical use.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
The description partially compensates for the 0% schema description coverage by explaining the folder_id parameter's behavior when omitted. However, it does not add meaning for local_path or name beyond their names, leaving some ambiguity.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states 'Upload a local file to Drive,' which is a specific verb+resource combination. It distinguishes from sibling tools like drive_search, drive_list, and drive_delete by focusing on the upload action.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
The description mentions 'Requires user confirmation before calling,' providing important usage context. However, it does not specify when not to use this tool or suggest alternatives, so it falls short of a perfect score.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Archive a message in Hermes's inbox (removes the INBOX label).
Never call without user confirmation.
Returns dict with key: id (the message ID that was archived).
| Name | Required | Description | Default |
|---|---|---|---|
| message_id | Yes |
| Name | Required | Description |
|---|---|---|
No output parameters | ||
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
Annotations show destructiveHint=false and idempotentHint=true, but the description adds that the operation removes the INBOX label and returns a dict with 'id'. This provides behavioral context beyond the annotations.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is very concise: two sentences plus a return note. It front-loads the main purpose and adds a critical usage warning. No wasted words.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
For a simple tool with one parameter and an output schema, the description covers the purpose, return value, and a usage caution. It could mention parameter details, but overall is sufficient given the tool's simplicity.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
With schema coverage at 0%, the description should compensate by explaining the message_id parameter, but it does not. No format, source, or details are provided beyond the schema itself.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states the verb 'Archive' and the resource 'a message in Hermes's inbox', and explains the effect of removing the INBOX label. This distinguishes it from sibling tools like mail_mark_read or mail_search.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
The description includes the warning 'Never call without user confirmation', but does not explicitly specify when to use this tool versus other mail-related sibling tools. It lacks a clear when-to/when-not-to guideline.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Fetch a single email by ID. Unwraps forwarded messages to extract the original.
Returns dict with keys: id, thread_id, sender (the user who
forwarded this message to Hermes — use as the to for replies),
original_sender, original_subject, original_body (plain text),
in_reply_to (Message-ID header for threading), forwarding_note
(text the user wrote above the forward delimiter, or null),
attachment_paths (list of local file paths in ~/.cache/hermes-google/).
| Name | Required | Description | Default |
|---|---|---|---|
| message_id | Yes |
| Name | Required | Description |
|---|---|---|
No output parameters | ||
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
Annotations already declare readOnlyHint=true, so safety is clear. Description adds valuable behavioral detail: unwraps forwarded messages to extract original. This goes beyond annotations. No contradiction.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
Concise two-paragraph structure. First sentence states core purpose. Second paragraph lists return keys clearly. No redundant information.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Given that an output schema exists (context signal), the description adequately explains the return values. However, it does not mention error handling (e.g., if ID not found). For a simple read tool, this is good but not perfect.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
Schema has 0% description coverage, yet the description does not explain what message_id is (e.g., 'Gmail message ID'). While it may be obvious, the description fails to add meaning beyond the parameter name. Baseline 3 is reduced because the schema lacks descriptions and the tool description does not compensate.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
Clearly states it fetches a single email by ID. Distinguishes from sibling tools like mail_search (which searches) and mail_list_pending (which lists pending). The verb 'Fetch' and resource 'email' are specific.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
Implied usage: when you have an email ID and want full details. But no explicit guidance on when not to use it or alternatives. For example, could mention that mail_search is for finding emails by query.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
List unread forwarded emails in Hermes's inbox, newest first.
Returns a list of dicts, each with keys: id, thread_id, sender,
subject, date, snippet.
| Name | Required | Description | Default |
|---|---|---|---|
| limit | No | Max results (1-100) |
| Name | Required | Description |
|---|---|---|
| result | Yes |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
Annotations already indicate readOnlyHint=true and destructiveHint=false. The description adds value by disclosing the return format (list of dicts with specific keys), which is beyond what annotations provide. It does not, however, mention any pagination or rate-limiting behavior.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is concise: two sentences that front-load the core purpose and provide essential return structure. No unnecessary words or repetition, and every sentence adds value.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Given the tool has an output schema (as per context signals), the description's explanation of return fields is redundant but harmless. The description, combined with schema and annotations, fully specifies the tool's behavior for an AI agent to select and invoke correctly.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
Schema coverage is 100% for the single optional parameter 'limit', which is fully described in the schema (default, min, max). The description does not add any additional semantic information about the parameter beyond what the schema already conveys.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states the tool lists unread forwarded emails in Hermes's inbox, sorted newest first. It uses specific verbs ('list') and resources ('unread forwarded emails'), distinguishing it from siblings like mail_search (broad search) and mail_get (single email retrieval).
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
The description implies the tool is for listing pending (unread forwarded) emails, but does not explicitly state when to use it versus alternatives like mail_search or mail_archive. No guidance on prerequisites or exclusion criteria is provided.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Mark a message as read in Hermes's inbox (removes the UNREAD label).
Returns dict with key: id (the message ID that was marked).
| Name | Required | Description | Default |
|---|---|---|---|
| message_id | Yes |
| Name | Required | Description |
|---|---|---|
No output parameters | ||
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
Description adds the behavioral context of removing the UNREAD label, which is not encoded in annotations (non-destructive, idempotent). No contradiction with annotations.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
Two concise sentences, front-loaded with primary action, clear and efficient.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Provides expected return format (dict with 'id'). Siblings suggest many mail tools, but no prerequisites or edge cases (e.g., already read) are discussed. Adequate for a simple idempotent operation.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
Single parameter 'message_id' has no description in schema (0% coverage) and description provides no additional semantics beyond its role as the message identifier.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
Description explicitly states the action: 'Mark a message as read in Hermes's inbox (removes the UNREAD label).' It clearly distinguishes from sibling tools like 'mail_archive' and 'mail_search'.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
Does not provide explicit guidance on when to use this tool versus alternatives. The description implies usage for marking a message read, but no when-not-to-use or comparison with siblings.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Search emails in Hermes's Gmail using Gmail search syntax.
Supports the same operators as the Gmail search bar (e.g., from:alice,
subject:invoice, after:2026/04/01).
Returns a list of dicts, each with keys: id, thread_id, sender,
subject, date, snippet.
| Name | Required | Description | Default |
|---|---|---|---|
| limit | No | Max results (1-100) | |
| query | Yes |
| Name | Required | Description |
|---|---|---|
| result | Yes |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
Annotations declare readOnlyHint=true, which aligns with the search nature. The description adds value by specifying the use of Gmail search syntax and listing returned fields. It does not mention pagination or rate limits, but the limit parameter covers result count. No contradictions with annotations.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is concise with three sentences, each serving a purpose: stating the function, explaining syntax, and describing the return value. No unnecessary text, and the key information is front-loaded.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
The description covers the search syntax, return format, and limit parameter. It mentions it searches Hermes's Gmail, providing context. However, it could mention ordering or that results are from a single mailbox. Overall, it is fairly complete for a search tool with an output schema.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
The schema has 50% description coverage; the query parameter lacks a description. The description compensates by providing examples of Gmail search operators (from:, subject:, after:), adding significant semantic meaning beyond the input schema.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states the tool searches emails in Hermes's Gmail using Gmail search syntax, specifying a verb and resource. It distinguishes from sibling tools like mail_get (single email retrieval) and mail_list_pending (listing pending emails) by focusing on search with complex queries.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
The description implies usage for search tasks but does not explicitly state when to use this tool over alternatives. No guidance on when not to use or comparison with mail_list_pending or other list tools is provided, leaving it to the agent to infer.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Send a draft email from Hermes's account.
The destination should be the sender returned by mail_get — i.e. the
user who forwarded the message to Hermes. To reply to a thread, pass the
in_reply_to value from mail_get.
Returns dict with key: id (sent message ID).
| Name | Required | Description | Default |
|---|---|---|---|
| to | Yes | ||
| body | Yes | ||
| subject | Yes | ||
| in_reply_to | No |
| Name | Required | Description |
|---|---|---|
No output parameters | ||
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
Annotations are sparse (no readOnlyHint, destructiveHint false). The description adds value by specifying the return value (`id`). It does not contradict annotations, but could mention if sending a draft is irreversible or if any side effects occur. Still, it provides useful behavioral context beyond annotations.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is only three sentences, each sentence serves a distinct purpose: purpose, usage, return value. It is front-loaded and contains no fluff.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Given the tool's simplicity (4 params, no nested objects), the description covers the core behavior, usage in context of other mail tools, and return format. It does not discuss error cases or prerequisites like draft existence, but it is adequate for the minimal complexity.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
Schema description coverage is 0%, so the description must compensate. It explains `to` and `in_reply_to` but does not describe `subject` or `body`. However, `subject` and `body` are self-explanatory. Partial coverage, but not fully compensating for all parameters.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states the verb ('Send'), resource ('draft email'), and origin ('from Hermes's account'). It differentiates from sibling tools like mail_get by specifically indicating sending drafts, and provides context for replying.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
The tool description tells when to use it (to send drafts) and provides specific guidance: use the `sender` from mail_get as destination, and pass `in_reply_to` for thread replies. It does not explicitly state when not to use it, but the context is clear enough.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Recent tool additions, removals, and schema changes observed during successful MCP inspections. Dates show when Glama detected each change.
v0.1.0auth_statuscal_create_eventcal_delete_eventcal_list_calendarscal_list_eventscal_update_eventdrive_deletedrive_getdrive_listdrive_movedrive_searchdrive_updatedrive_uploadmail_archivemail_getmail_list_pendingmail_mark_readmail_searchmail_send_draftEach tool targets a distinct operation within its service (auth, calendar, drive, mail). Tools like cal_create_event and cal_delete_event have clearly different purposes, and no two tools overlap in functionality.
All tool names follow a consistent verb_noun pattern with service-specific prefixes (auth_, cal_, drive_, mail_). Verbs like list_, create_, delete_, update_ are used uniformly, making it easy to predict tool behavior.
With 19 tools covering four Google services (auth, calendar, drive, mail), the count is well-scoped. Each service has a natural number of tools (5 calendar, 7 drive, 6 mail, 1 auth) without redundancy or gaps.
The tool surface provides CRUD operations for calendar events and drive files, plus comprehensive email search and management. Missing features like sending arbitrary emails or managing labels are minor given the server's focus on a personal assistant's inbox, but could be considered a gap for full email coverage.
Unclaimed servers have limited discoverability.
If you are the server author, to access and configure the admin panel.
Multiple Google accounts (Gmail, Calendar, Drive, Contacts, Tasks) in one Claude connector.
Give Claude only the Google Drive files you choose. Every action logged.
Permissioned access to Gmail, Drive and Calendar via the user's own Google account
Personal CRM for Claude. Contacts live as plain-text files in your own Google Drive.
We provide all the information about MCP servers via our MCP API.
curl -X GET 'https://glama.ai/api/mcp/v1/servers/jimmy-larsson/hermes-google'
If you have feedback or need assistance with the MCP directory API, please join our Discord server
