Use Cadreen with MCP

Model Context Protocol (MCP) lets any MCP-compatible client — Cursor, Claude Desktop, Continue, and others — use Cadreen as a tool provider. You get governance, memory, and learning on top of your existing tools.

Two ways to connect

A. Direct via SSE (zero install)

Point any MCP client at Cadreen's SSE endpoint. No local packages needed.

B. Install the stdio server (local)

Install @cadreen/mcp globally. Runs locally over stdio.

A. Connect directly via SSE

Point any MCP-compatible client at Cadreen's SSE endpoint. No install required — just configure and go.

SSE URLhttps://accomplishanything.today/api/v1/cadreen/mcp/sse

AuthAuthorization: Bearer sk_cadreen_...

ProtocolJSON-RPC 2.0 over SSE

Claude Desktop config (SSE)

JSON

{
  "mcpServers": {
    "cadreen": {
      "url": "https://accomplishanything.today/api/v1/cadreen/mcp/sse",
      "transport": "sse"
    }
  }
}

B. Install the stdio MCP server

Install the MCP server locally. Runs as a stdio process — no network configuration needed.

TEXT

npm install -g @cadreen/mcp

Claude Desktop config (stdio)

JSON

{
  "mcpServers": {
    "cadreen": {
      "command": "cadreen-mcp",
      "env": {
        "CADREEN_API_KEY": "sk_cadreen_..."
      }
    }
  }
}

Run directly

TEXT

# Run the server
cadreen-mcp

# Or with npx
npx @cadreen/mcp

C. Register external MCP servers for Cadreen

This goes the other direction — Cadreen connects to your MCP servers and uses their tools. Cadreen's governance layer applies to every tool call.

Register via API

BASH

curl -X POST https://accomplishanything.today/api/v1/cadreen/connections/mcp \
  -H "Authorization: Bearer sk_cadreen_..." \
  -H "Content-Type: application/json" \
  -d '{
    "name": "My MCP Server",
    "transport": "sse",
    "url": "https://your-mcp-server.example.com/sse"
  }'

SDK examples

Python

TEXT

from cadreen import Cadreen

client = Cadreen(api_key="sk_cadreen_...")

# Register your MCP server
client.connections.register_mcp(
    name="My MCP Server",
    transport="sse",
    url="https://your-mcp-server.example.com/sse"
)

TypeScript

TEXT

import { Cadreen } from "@cadreen/sdk";

const client = new Cadreen({ apiKey: "sk_cadreen_..." });

// Register your MCP server
await client.connections.registerMCP({
  name: "My MCP Server",
  transport: "sse",
  url: "https://your-mcp-server.example.com/sse",
});

TEXT

package main

import (
    "context"
    "github.com/timothy-billingrails/cadreen-sdks/go/cadreen"
)

func main() {
    client := cadreen.NewClient("sk_cadreen_...")

    // Register your MCP server
    client.Connections.RegisterMCP(context.Background(), cadreen.RegisterMCPParams{
        Name:      "My MCP Server",
        Transport: "sse",
        URL:       "https://your-mcp-server.example.com/sse",
    })
}

All 22 tools

When you connect via MCP (A or B), these capabilities become available as tools:

Intent

cadreen_intentSend a request. Describe what you want done.

Agents

cadreen_agents_listList all agents in your workspace.

cadreen_agents_createCreate a new agent.

cadreen_agents_getGet agent details, status, and health.

Knowledge

cadreen_knowledge_searchSearch an agent's knowledge base.

cadreen_knowledge_addTeach an agent something new.

Governance

cadreen_governance_listList governance policies.

cadreen_governance_createCreate a governance policy.

Federation

cadreen_federation_listList federation links.

cadreen_federation_createCreate a federation link.

Responses

cadreen_responses_createCreate an OpenAI-compatible response.

External Agents (A2A)

cadreen_external_agents_listList external A2A connections for an agent.

cadreen_external_agents_list_allList all external connections across workspace.

cadreen_external_agents_connectConnect to an external agent via Agent Card URL.

cadreen_external_agents_getGet connection details, status, and capabilities.

cadreen_external_agents_approveApprove a pending connection.

cadreen_external_agents_suspendSuspend an active connection.

cadreen_external_agents_revokeRevoke a connection permanently.

cadreen_external_agents_deleteDelete a connection.

cadreen_external_agents_list_interactionsList tasks sent/received for a connection.

cadreen_external_agents_get_settingsGet workspace external agent settings.

cadreen_external_agents_update_settingsEnable/disable external agents.

When to use MCP vs other methods

Direct API

You want Cadreen as your model provider

Same OpenAI-compatible endpoint, full intelligence layer.

MCP

You want Cadreen as a tool inside another AI

Cadreen's governance and memory available as tools in Cursor, Claude Desktop, etc.

A2A

You want Cadreen to talk to another agent

Cross-platform agent collaboration with governance.

Note

MCP is best when you want Cadreen's governance and memory available as tools inside another AI system. For direct model access, use the OpenAI-compatible endpoint.

Next steps:

Connect as model provider External agents (A2A)Set up rules