Click on "Deploy 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., "@Capacities MCP Bridge (Unofficial) save this link to my daily notes: https://example.com"
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.
This project provides several methods for connecting to the Capacities note taking app API using the Model Context Protocol (MCP). It was 100% vibe-coded with the help of Gemini 2.5 Pro. While it works, it relies on third-party adapter services and, in some cases, your own hosting. Use at your own risk.
For everyone looking for a native Bun implementation, there exists another project https://github.com/jem-computer/capacities-mcp
Related MCP server: Fastidious MCP Server
Setup for Claude Desktop (with Node.js Bridge)
Use this if you need to connect Claude Desktop and it requires simple tool names.
Setup for Simple SSE Clients (CLion, VS Code)
Use this for a direct connection from a compatible IDE plugin.
Advanced Setup for Genspark (Self-Hosted Server)
The most reliable and robust method. Use this for Genspark or any other client if you have your own server/VPS.
A Capacities API Token. You can generate this from your Capacities account settings if you have a paid account.
The MCP Link Generator tool: https://mcp-link.vercel.app/
This method uses the included capacities-claude-bridge.js script to act as a translator between Claude Desktop (which needs simple tool names) and the MCP adapter (which creates complex names).
Install Node.js: Go to the official Node.js website and download and install the LTS version for your operating system.
Download the Bridge Code: Open a terminal (like Git Bash, Command Prompt, or PowerShell) and run the following commands:
# Clone the repository to a permanent location
git clone https://github.com/natkitten/capacities-mcp-bridge-unofficial.git
# Navigate into the project folder
cd capacities-mcp-bridge-unofficialInstall Dependencies: While inside the capacities-mcp-bridge-unofficial folder, run:
npm installGo to mcp-link.vercel.app.
Fill in the form:
OpenAPI Specification URL:
Use this specific URL for the Claude bridge setup. It uses snake_case operationIds.
https://gist.githubusercontent.com/natkitten/e6ce1335c2cdad87a9237156c5cda315/raw/capacities_openapi_2.jsonAPI Base URL: https://api.capacities.io
HTTP Headers: Enter your Authorization header:
Authorization: Bearer YOUR_CAPACITIES_API_TOKENPath Filters: Leave this field completely empty.
Encoding Options: Select "Base64 (JSON Encoded)".
Click "Generate MCP Link" and copy the resulting URL.
Open the capacities-claude-bridge.js file (located in the folder you just downloaded) in a text editor.
Paste the URL you just copied, replacing the placeholder text for the SSE_URL constant.
const SSE_URL = 'https://mcp-openapi-to-mcp-adapter.onrender.com/sse?code=...'; // YOUR URL HERESave the file.
Open File Explorer and navigate to your Claude Desktop config file by pasting this path into the address bar:
%APPDATA%\Claude Desktop\claude_desktop_config.json
Open the file and add the mcp_bridges section as shown below.
{
"mcpServers": {
"capacities": {
"command": "node",
"args": ["C:\\path\\to\\your\\capacities-mcp-bridge-unofficial\\capacities-claude-bridge.js"]
}
}
}CRUCIAL: Replace C:\\path\\to\\your\\capacities-mcp-bridge-unofficial with the actual, absolute path to the folder where you cloned the repository. Remember to use double backslashes \\.
Save the claude_desktop_config.json file and restart Claude Desktop.
Endpoint | What it does | Typical use-case |
| Lists all Capacities spaces the token can access | Show a picker or verify the token |
| Returns structures, collections & property definitions of a space | Needed once at startup to map IDs to human labels |
| Full-text or title search across one or many spaces | Let the LLM find existing notes before it creates new ones |
| Saves an external URL (and optional tags/markdown) into a space | Quick bookmarking from chat |
| Appends Markdown to today’s daily note in a space | Fast journaling / meeting-note dump |
Current rate-limits (per user / 60 s window):
/spaces & /space-info: 5 requests
/search: 120 requests
/save-weblink: 10 requests
/save-to-daily-note: 5 requests
For everything else (errors, structures, OpenAPI spec), see the official docs ➜ https://api.capacities.io/docs/.
https://github.com/user-attachments/assets/257aada5-5f65-4aec-b7da-4766452e3cb5
This method is for MCP plugins that can handle the mcplink_... tool names directly. No bridge script is needed.
Follow Step 1.2 above to generate your unique Server URL from mcp-link.vercel.app, using the capacities_openapi_2.json and leaving "Path Filters" empty.
Find the MCP configuration file for your IDE. It's often located at YOUR_HOME_DIRECTORY/.mcp/servers.json.
Add the following entry, replacing the placeholder with your generated URL:
{
"servers": {
"capacities": {
"url": "PASTE_YOUR_GENERATED_URL_HERE"
}
}
}Restart your IDE. The tools should appear with their full mcplink_... names.
This is the most reliable method. It runs the MCP server on your own VPS. This setup requires a slightly different OpenAPI spec to work around a parser bug in the self-hosted server package.
Connect to your VPS via SSH (ssh root@YOUR_VPS_IP).
Install nvm (Node Version Manager):
curl -o- https://raw.githubusercontent.com/nvm-sh/nvm/v0.39.7/install.sh | bashActivate nvm:
export NVM_DIR="$HOME/.nvm"
[ -s "$NVM_DIR/nvm.sh" ] && \. "$NVM_DIR/nvm.sh"Add nvm to your shell profile so it loads automatically on every login:
echo 'export NVM_DIR="$HOME/.nvm"' >> ~/.bashrc
echo '[ -s "$NVM_DIR/nvm.sh" ] && \. "$NVM_DIR/nvm.sh"' >> ~/.bashrcInstall Node.js (LTS version):
nvm install --ltsCreate a project directory on your VPS and navigate into it:
mkdir capacities-mcp-server
cd capacities-mcp-serverInitialize a Node.js project: This creates the package.json file.
npm init -ySet the project type to "module": Open the package.json with nano package.json and add "type": "module", after the "main": "index.js", line.
Install pm2 and the server package locally: We install them here to keep the project self-contained.
npm install pm2 @ivotoby/openapi-mcp-serverCreate an ecosystem.config.cjs file for pm2. The .cjs extension is important.
nano ecosystem.config.cjsPaste the following configuration into the file. This uses the locally installed pm2 and server script.
module.exports = {
apps : [{
name : 'capacities-mcp',
script : './node_modules/@ivotoby/openapi-mcp-server/dist/cli.js',
args : [
'--openapi-spec',
'https://gist.githubusercontent.com/natkitten/37e88b5dab4195b0f4d650f31f5505bf/raw/capacities_openapi_3.json',
'--api-base-url',
'https://api.capacities.io',
'--headers',
'Authorization:Bearer YOUR_CAPACITIES_API_TOKEN',
'--toolNameFormat',
'**',
'--transport',
'http',
'--host',
'0.0.0.0',
'--port',
'8448' // Or your preferred port
]
}]
}Replace YOUR_CAPACITIES_API_TOKEN with your actual token.
Save and exit (CTRL + X, Y, Enter).
Start the server using the local pm2:
./node_modules/pm2/bin/pm2 start ecosystem.config.cjsCheck logs to confirm it's listening on your port: pm2 logs capacities-mcp
Save the configuration for reboots: pm2 save
Enable pm2 on startup: Run pm2 startup and follow the on-screen instructions.
Firewall: Open your chosen port (8448 in this example) on both your VPS firewall (sudo ufw allow 8448) and your cloud provider's firewall (in the Hostinger dashboard).
Caddy: For a secure HTTPS URL, set up a reverse proxy in your Caddyfile.
mcp.yourdomain.com {
reverse_proxy localhost:8448
}Reload Caddy to apply the changes (sudo systemctl reload caddy).
Genspark:
Server Type: StreamableHttp
Server URL: https://mcp.yourdomain.com/mcp (using the /mcp path).
Request Header:
{"Content-Type": "application/json"}Add the server and test it.
Licensed under the MIT License — see LICENSE for details
Get structures and collections of a space
| Name | Required | Description | Default |
|---|---|---|---|
| searchParams | No |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
No annotations are provided, so the description must disclose behavioral traits. It only says 'Get', implying a read operation, but does not elaborate on whether any side effects exist, required permissions, error behavior, or what 'structures and collections' entail. This is insufficient for a tool with no annotation support.
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 a single, clean sentence that is front-loaded with the key action. It avoids unnecessary words and is easy to parse. While more detail could be added, the conciseness is appropriate for the limited content it provides.
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 (one parameter, no output schema, no annotations), the description gives a basic idea that the tool returns structures and collections. However, it lacks details on the exact return format, any pagination, error conditions, or what 'structures' and 'collections' mean in this context. It is barely adequate for a minimal understanding.
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 zero description coverage, and the description does not compensate by explaining the spaceid parameter. It only says 'of a space', which weakly hints at the need for a space identifier but does not clarify the parameter format, nesting, or purpose. The parameter name 'spaceid' partially helps, but the description adds minimal value 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 'Get structures and collections of a space' clearly states a specific verb (Get) and resource (structures and collections of a space), making it evident what the tool does. It does not explicitly differentiate from sibling tools like get_spaces, but the phrase 'of a space' implies it targets a single space's details rather than listing spaces.
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?
There is no guidance on when to use this tool versus alternatives such as get_spaces or search_content. The description does not state any prerequisites, exclusions, or context in which this tool should be preferred, leaving the agent without explicit selection criteria.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Get your spaces
| Name | Required | Description | Default |
|---|---|---|---|
No parameters | |||
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
No annotations are provided, so the description must fully disclose behavior. It only states 'Get your spaces' with no mention of authentication, rate limits, pagination, or return format. The verb 'get' implies read-only, but no additional behavioral context is given, making this a minimal disclosure.
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 a single four-word phrase with no wasted language. It is front-loaded and appropriately sized for a tool with no parameters, though it lacks any supplemental details.
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 parameterless list tool, the description provides the essential meaning: retrieving spaces. However, it omits any context about the structure of returned data or how it relates to get_space_info, leaving some ambiguity for a complex domain.
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?
This tool has zero parameters, so the input schema fully covers the parameter space. The description adds nothing about parameters, but none are needed, so the baseline of 4 applies.
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 uses the specific verb 'get' and resource 'spaces', clearly indicating a retrieval operation. However, it does not distinguish from sibling tool 'get_space_info', which likely focuses on a single space's details, so it doesn't earn a 5.
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 retrieving the user's spaces but provides no explicit guidance on when to use this instead of get_space_info or other siblings. There are no stated alternatives or exclusions, so it stays at an implied level.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Save text to today's daily note
| Name | Required | Description | Default |
|---|---|---|---|
| requestBody | No |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
With no annotations provided, the description must fully disclose behavioral traits. It only states the basic action and does not explain whether content is appended or overwritten, what permissions are required, or any side effects. This is a significant gap for a mutation tool.
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 a single concise sentence with no filler, which is structurally efficient. However, it is under-specified, missing essential details about parameters and behavior, so the brevity is not a virtue in this context.
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 moderate complexity (nested requestBody, no output schema, no annotations), the description is far from complete. It does not mention that spaceId and mdText are required, nor describe what the tool returns, making it inadequate for safe invocation.
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 one nested requestBody with fields like spaceId and mdText, but the description does not explain any of them. With 0% schema description coverage, the description completely fails to compensate, leaving parameter meanings ambiguous.
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 'Save text to today's daily note' clearly identifies the action (save text) and the resource (today's daily note), which distinguishes it from sibling tools like save_weblink (saving URLs) and search_content. However, it lacks elaboration on what 'text' means or the note's context, but the core purpose is unambiguous.
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 guidance on when to use this tool versus alternatives such as save_weblink or search_content. There is no mention of prerequisites, preferred use cases, or exclusions, leaving the agent without direction on selection.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Save a weblink to a space
| Name | Required | Description | Default |
|---|---|---|---|
| requestBody | No |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
No annotations are provided, so the description must carry the full burden of behavioral disclosure. It only says 'save a weblink to a space', implying a mutation but providing no details about required permissions, side effects, errors, response behavior, or what happens to existing data. This is insufficient for a write operation.
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 a single concise sentence with no wasteful words, and it is front-loaded with the verb. However, it is so brief that it lacks structure or elaboration, making it more under-specified than appropriately 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 nested requestBody with six properties, no annotations, and no output schema, the description is far too minimal. It provides no information about required fields (e.g., spaceId, url), optional enhancements (tags, mdText), or what the response looks like, leaving the agent without enough context to invoke the tool 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 description coverage is 0%, and the description does not name any parameters. It vaguely maps 'weblink' to the url field and 'space' to spaceId, but completely ignores tags, mdText, titleOverwrite, descriptionOverwrite, and the nested requestBody structure. An agent would have to guess at the meaning of these fields from the schema alone.
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 uses a specific verb ('save'), resource ('weblink'), and destination ('space'), which makes the core purpose clear. It is distinguishable from sibling tools like 'get_spaces' and 'save_to_daily_note' due to the explicit target, but it does not explicitly name alternatives or edge cases.
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?
There is no guidance on when to use this tool vs. alternatives like 'save_to_daily_note', no prerequisites, no context about typical scenarios, and no mention of limitations. The description simply states what it does, not when to choose it.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Search for content
| Name | Required | Description | Default |
|---|---|---|---|
| requestBody | No |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
With no annotations available, the description must disclose behavioral traits, but it only says 'search' without explaining return format, filtering behavior, pagination, limitations, or side effects. The nested schema hints at advanced capabilities, but the description offers no behavioral context beyond the name.
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 short, but the single phrase 'Search for content' is under-specified rather than appropriately concise. It does not earn its place because it adds no information beyond the tool name, making it a case of under-specification, not efficiency.
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 nested requestBody schema, enum values, required fields, and lack of output schema or annotations, the description is drastically incomplete. It provides no context about how the tool fits into the broader workspace of sibling tools or what an agent should expect when invoking it.
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 description coverage is 0%, so the description must compensate by explaining the requestBody fields (searchTerm, spaceIds, mode, filterStructureIds). It provides no such explanation, leaving the agent to infer parameter semantics solely from schema structure, which is insufficient for correct invocation.
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 'Search for content' essentially restates the tool name without adding specificity about what content is searched, in what scope, or how it differs from sibling tools like get_spaces or get_space_info. It is a tautological expression of the name, not a meaningful clarification of purpose.
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 guidance is provided on when to use this tool versus alternatives such as get_spaces or get_space_info. The description does not mention intended use cases, prerequisites, or conditions under which search_content is preferred. There is no misleading information, but also no direction.
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.
v1.0.0get_space_infoget_spacessave_to_daily_notesave_weblinksearch_contentScored across 5 tools
Each tool targets a distinct action and resource: list spaces, get space details, save text to daily note, save weblink, and search. There is no overlap or ambiguity between tool purposes.
All tool names follow a consistent verb_noun pattern with lowercase and underscores (e.g., get_spaces, save_to_daily_note, search_content). The naming style is uniform and predictable.
5 tools is well-scoped for a lightweight bridge. It provides core read, write, and search capabilities without unnecessary redundancy or bloat.
The toolset covers essential workflows: retrieving spaces, saving content, and searching. Minor gaps exist such as lack of general note creation or update/delete operations, but these are acceptable for an unofficial bridge.
MCP-native notes and memory for ChatGPT, Claude, and other AI tools.
Persistent memory layer for AI tools. Save and recall notes across Claude and other MCP clients.
Connect AI to your flomo notes. Search, create, edit notes and manage tags via MCP.
Search, read, and write your Apple Notes from ChatGPT/Claude via a local Mac agent + MCP relay.
