- Category
- AI Agents
- Rank
- No. 2112Tools index
Previous survey · No. 2107 ·
- Pricing
- Open Source
- Type
- TOOL
- Builder
- collaborator-ai
- GitHub
- 2 stars
- Date
About
Invoke your OpenClaw agent persona in any AI tool via an MCP server.
What it does
Collab MCP turns an OpenClaw workspace into shared context for compatible coding assistants. It loads identity, user information, memory, recent notes, and conversation history, supports topic recall, and can write decisions or observations back to memory. Markdown files remain the source of truth, with optional Seedvault search.
Why it's ranked here
Its appeal is a small, understandable architecture: local files, a Node server, and standard MCP transports. It preserves useful context across coding tools without requiring a database or account. The rough edges matter, though. Authentication is basic, remote deployment needs extra infrastructure, and configuration guidance does not fully match the supplied source.
What's good
The tool covers a full context loop rather than simple retrieval. It can bootstrap a working identity, search for focused background, and record new learning. It supports standard input and output, server-sent events, and streamable HTTP, so several MCP clients can connect. Local markdown storage keeps the agent's accumulated context inspectable and portable.
Tradeoffs
This is tightly coupled to OpenClaw's workspace structure and naming conventions. The server must remain running, and restarts require separate process supervision. URL-based API keys can leak through logs or history, while the OAuth flow auto-approves without a login page. The documented workspace variable is not used by the supplied source, which instead derives the workspace from the OpenClaw root.
How to use it well
Use it when an OpenClaw agent already holds durable project context and you want that context available inside multiple coding assistants. Start sessions by loading the full context, recall details only when needed, and record durable decisions afterward. Keep it local unless you can provide HTTPS and careful token handling. It does not replace general search, cloud synchronization, process supervision, or a standalone memory system.
Technical notes+
package.json defines an ES module TypeScript package with a CLI binary, build, start, and watch scripts, plus the MCP SDK and Zod as runtime dependencies. tsconfig.json targets ES2022, uses Node16 module resolution, enables strict checking, and compiles src/**/* into a build directory. In the supplied portion of src/index.ts, the server imports SSE, Streamable HTTP, and stdio transports; reads workspace markdown and OpenClaw session JSONL; caps recent conversation context by message and word counts; optionally queries Seedvault; and accepts OPENCLAW_AUTH_TOKEN or COLLAB_AUTH_TOKEN. A notable mismatch is that README.md and SKILL.md document OPENCLAW_WORKSPACE, while src/index.ts derives WORKSPACE from OPENCLAW_ROOT. The source file is truncated, so the complete recall, record, routing, and authentication implementations cannot be verified from the provided text. collab.md and skills/collab/SKILL.md provide client-side guidance for loading the bootstrap context first.
Observed
- License
- MIT
- Primary language
- TypeScript
- Packaging
- Node.js ES module package with an npm CLI binary and TypeScript build
- Install surface
- Clone the repository, install npm dependencies, compile TypeScript, and run the server
- Interfaces
- MCP server over stdio, server-sent events, and Streamable HTTP
- Platform support
- Documented for Claude Code, Codex, Cursor, Claude Desktop, and other MCP-compatible clients
- Storage model
- Local OpenClaw markdown workspace and session JSONL files, with optional Seedvault search
Read from README.md, package.json, src/index.ts, SKILL.md, collab.md, tsconfig.json, skills/collab/SKILL.md.
What it can do
Invoke OpenClaw agent persona
AI tool session → Active OpenClaw agent instance
Establish MCP server connection
Client application request → MCP protocol connection
Deploy agent persona across multiple AI tools
OpenClaw agent configuration → Consistent agent behavior in different AI environments
Enable cross-platform agent collaboration
Multiple AI tool instances → Synchronized agent responses
Tags
Tech Stack
Comments (0)
No comments yet
Editorially curated, with community endorsements as a secondary signal. Corrections welcome.