atlas-introspect-mcp
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
No arguments | |||
Instructions
Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.
This server publishes no instructions, or was last inspected before Glama recorded them.
Capabilities
Features and capabilities supported by this server
Protocol revision2025-11-25
| Capability | Details |
|---|---|
| tools | {
"listChanged": true
} |
| logging | {} |
| prompts | {
"listChanged": false
} |
| resources | {
"subscribe": false,
"listChanged": false
} |
| extensions | {
"io.modelcontextprotocol/ui": {}
} |
| experimental | {} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| atlas_statusA | Check that ATLAS is up and report what this server is pointed at. Start here when another tool fails: it distinguishes "ATLAS is down" from "the feature is disabled" from "this server is aimed at the wrong URL". |
| list_conversationsA | List saved conversations, newest first. Returns id, title, model, timestamps, message count, and a short preview.
Use |
| search_conversationsC | Search conversations by title and message content. |
| get_conversationA | Read one full conversation, including its messages. Long messages are clipped per-message so a single large tool result cannot
swamp the answer; raise |
| list_conversation_tagsA | List the tags this user has applied to conversations. |
| get_mcp_configA | Read config/mcp.json — the whole file, or one server entry. |
| upsert_mcp_serverA | Add or replace one server entry in config/mcp.json. The entry is replaced wholesale, not merged — read the current entry with
A stdio entry looks like:: |
| delete_mcp_serverA | Remove a server entry from config/mcp.json (previous file is backed up). |
| replace_mcp_configA | Overwrite config/mcp.json with a complete JSON document. For restructuring several entries at once. Rejected unless |
| list_workspacesA | List this user's workspaces (saved bundles of prompt/tool/RAG selections). |
| create_workspaceB | Create a workspace. The selection fields are exactly what ATLAS restores when the workspace is applied; unknown keys are rejected by the API, so use these arguments rather than a free-form config blob. |
| update_workspaceA | Update a workspace's name, description, or selections.
|
| delete_workspaceA | Delete a workspace. Not recoverable — confirm with the user first. |
| list_custom_promptsB | List this user's custom prompt library. |
| get_custom_promptA | Read one custom prompt in full, by id. |
| create_custom_promptB | Add a prompt to this user's library. |
| update_custom_promptA | Edit a custom prompt's title or content.
|
| delete_custom_promptA | Delete a custom prompt. Not recoverable — confirm with the user first. |
| list_prompt_filesA | List the markdown prompt files ATLAS loads from disk. These are the defaults (system prompt, agent reason/observe/summary prompts), separate from the per-user library above. Editing one changes behaviour for every chat, and takes effect on the next ATLAS restart. |
| read_prompt_fileC | Read one on-disk prompt file. |
| write_prompt_fileA | Overwrite an on-disk prompt file, backing up the previous version. Affects every chat in this instance and applies on the next ATLAS restart.
Confirm with the user before changing |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
No prompts | |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
No resources | |
TDQS
Scored across 21 tools
Each tool targets a distinct resource and action: conversations, workspaces, custom prompts, prompt files, and MCP config are cleanly separated. Even similar operations like search_conversations vs list_conversations are differentiated by their descriptions (search vs browse). No two tools appear to do the same thing.
All tool names follow a consistent verb_noun pattern (list_, get_, create_, update_, delete_, read_, write_, search_, upsert_, replace_) with snake_case throughout. Only 'atlas_status' deviates slightly, but it's a special health-check tool and still fits the imperative style.
At 21 tools, the server is on the heavier side, but the scope (managing conversations, workspaces, prompts, prompt files, and MCP config) justifies the count. Each tool addresses a specific need, and there is no obvious duplication or bloat.
Workspaces, custom prompts, and MCP config all have full CRUD (or equivalent) coverage. Conversations lack create/delete, and tags are read-only, but those may be intentionally absent for an introspection tool. The ability to read and write prompt files covers the on-disk prompt management need. Minor gaps exist, but agents can work around them.