mcpField
R
Raccha AI

Raccha AI

VerifiedFeaturedRemote
raccha.aiDeveloper Tools0Updated 2w ago
RepositoryVisit website

About Raccha AI

Raccha AI is a remote MCP toolbox for agents, with namespaced storage, queues, authentication, and stateless utilities. It connects directly from MCP clients and uses a human-approved device flow to provision credentials.

Key features

  1. 01Namespaced key-value storage
  2. 02FIFO queue
  3. 03Client registration and device authorization
  4. 04JWT decoding and hashing
  5. 05Certificate inspection and CIDR calculations

Use cases

  1. 01Store and retrieve namespaced values
  2. 02Queue work in FIFO order
  3. 03Authorize a client through human approval
  4. 04Decode JWTs or inspect certificates

MCP support

Exposes the Raccha AI toolbox as a remote MCP endpoint over Streamable HTTP. Most tools require an owner or access credential; stateless utilities are exempt.

Configuration

Remote servers need no local install. Pick your client, then copy the config or install in one click.

Endpointhttps://raccha.ai/mcp
Copy install command
Claude CodeTerminal
Cursor.cursor/mcp.json
VS Code.vscode/mcp.json
Claude Desktopclaude_desktop_config.json
Windsurf~/.codeium/windsurf/mcp_config.json

Run commands in a terminal; paste JSON into the client’s MCP config and restart. Servers that need sign-in will prompt on first connect.

Tools

67
ParameterTypeRequiredDescription
pemstringYesA PEM-encoded X.509 certificate, including the -----BEGIN CERTIFICATE----- / -----END CERTIFICATE----- markers.

Details

From the project README

raccha.ai MCP server

MCP-first toolbox for agents: namespaced key-value storage, a FIFO queue, auth, and a handful of stateless utility tools (JWT decode, hashing, cert inspection, CIDR math). Free while in early access.

This repository is a manifest and connection guide only. raccha.ai's server source is closed; this repo exists so directories and registries that crawl GitHub can discover and list a server that otherwise has nothing to crawl. Everything here is real: the endpoint, the auth flow, the tool names. Nothing here is placeholder.

Discover

The server publishes a pre-standard .well-known/mcp discovery document:

GET https://raccha.ai/.well-known/mcp

It returns the same metadata as the Official MCP Registry server.json, including the streamable HTTP endpoint below. Clients can connect without hard-coding URLs.

Connect

The server is a real remote MCP endpoint, not a local package:

https://raccha.ai/mcp

Transport: streamable-http. Point any MCP-capable client at that URL directly, no install required.

Authenticate

Every tool call except the stateless utilities (jwt_decode, hash, cert_inspect, ip_cidr) requires a credential: an owner_key or access_key, passed as Authorization: Bearer <key> on the underlying HTTP call the MCP transport sits on.

If you're an MCP client connecting for the first time with no key, the recommended path is:

  1. Call the register_client tool with a client_name (OAuth 2.0 Dynamic Client Registration, RFC 7591). Returns a client_id. Grants nothing on its own.
  2. Call device_start with that client_id. Returns a short user code and a URL.
  3. A human opens that URL (https://raccha.ai/device.html) and approves, choosing either a full owner_key or an access_key scoped to specific roles. This step is unchanged and mandatory; dynamic client registration only removes the manual client-provisioning step, it never skips human approval.
  4. Call device_claim to pick up the credential once approved.

Full detail: https://raccha.ai/llms/auth.md.

Tools

Every HTTP route is also a generated MCP tool, built from the same OpenAPI contract the REST API is generated from; the two surfaces cannot drift from each other. Tool names and schemas match the operation IDs documented at https://raccha.ai/docs/ (Swagger UI over the real spec). Call tools/list against the endpoint above for the live, current set.

About the Dockerfile in this repo

It does not build raccha.ai's server. It builds a thin nginx reverse proxy onto the real, live endpoint at https://raccha.ai/mcp, for directories that require a runnable container to verify a server responds to MCP introspection. The actual service source stays closed.

For Kimi / Kimi Code users

A zero-info onboarding skill lives in this repo at .kimi-code/skills/raccha-mcp/SKILL.md. It tells an agent how to discover the endpoint, run the DCR + device-code flow, and store the resulting credential without asking the user for URLs or pre-shared secrets.

More

Comments

Markdown supported

New to MCP?

From what the protocol is to installing your first server, these answers cover most first questions.

An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.