@cohesivity/mcp

@cohesivity/mcp

Standalone Cohesivity MCP server over stdio:

npx -y @cohesivity/mcp

The server exposes the same local project-bootstrap interface shipped inside the Cohesivity plugin, packaged as a single npm bin so any MCP client can invoke it directly without the plugin installer. It requires no OAuth and runs entirely over stdio.

Cohesivity's hosted MCP server is separate: connect remote clients to https://cohesivity.ai/mcp. The former https://cohesivity.ai/mcp/manage endpoint is retired and returns HTTP 410.

Tools

Mutating tools require literal confirmed: true only when the current user request explicitly authorizes the exact action.

MCP client configuration

{
  "mcpServers": {
    "cohesivity": {
      "command": "npx",
      "args": ["-y", "@cohesivity/mcp"]
    }
  }
}

Attribution

create_tenant runs the Cohesivity quickstart, which names the agent that launched this server by walking its own parent processes (for example claude or cursor) and records that name with the new tenant. To tag tenants created through your integration, give the server a label. The tenant is then recorded as <agent>.<label>, for example claude.raven.

Pass it as a server argument:

{
  "mcpServers": {
    "cohesivity": {
      "command": "npx",
      "args": ["-y", "@cohesivity/mcp", "--attribution", "raven"]
    }
  }
}

or set COHESIVITY_ATTRIBUTION in the server's environment:

{
  "mcpServers": {
    "cohesivity": {
      "command": "npx",
      "args": ["-y", "@cohesivity/mcp"],
      "env": { "COHESIVITY_ATTRIBUTION": "raven" }
    }
  }
}

Use the argument form for catalogs that can pass args but can't set env. The server reads the label once at start. If both are set, a valid argument wins. The value must be 1-40 lowercase letters, digits, dots, underscores, or hyphens, starting with a letter or digit. An invalid value is ignored with a warning on stderr, and tenant creation proceeds without it.

Microsoft Agent Framework

Microsoft Agent Framework agents can use the hosted server as an MCP tool. When a task needs a backend, the agent creates a tenant and provisions what it needs, such as Postgres, Redis or object storage. No Cohesivity account or API key is required to start.

Python, after pip install agent-framework:

import asyncio

from agent_framework import Agent, MCPStreamableHTTPTool
from agent_framework.openai import OpenAIChatClient

async def main() -> None:
    async with Agent(
        client=OpenAIChatClient(model="gpt-5.4-mini"),
        tools=MCPStreamableHTTPTool(name="cohesivity", url="https://cohesivity.ai/mcp"),
    ) as agent:
        result = await agent.run("Set up a Postgres database for my todo app.")
        print(result.text)

asyncio.run(main())

.NET, after dotnet add package Microsoft.Agents.AI.OpenAI and dotnet add package ModelContextProtocol:

using Microsoft.Agents.AI;
using Microsoft.Extensions.AI;
using ModelContextProtocol.Client;
using OpenAI;
using OpenAI.Chat;

await using var mcpClient = await McpClient.CreateAsync(new HttpClientTransport(new()
{
    Endpoint = new Uri("https://cohesivity.ai/mcp"),
}));
var tools = await mcpClient.ListToolsAsync();

AIAgent agent = new OpenAIClient(Environment.GetEnvironmentVariable("OPENAI_API_KEY"))
    .GetChatClient("gpt-5.4-mini")
    .AsAIAgent(tools: [.. tools.Cast<AITool>()]);

Console.WriteLine(await agent.RunAsync("Set up a Postgres database for my todo app."));

Both examples use OpenAI and read OPENAI_API_KEY. Any other Agent Framework chat client works too. New tenants expire after 72 hours unless claimed; claim_tenant returns a link where the user approves keeping it.

Releasing

Releases publish to npm automatically via GitHub Actions when a version tag is pushed. The workflow uses npm Trusted Publishing (OIDC) — no NPM_TOKEN secret needed.

# bump version in package.json, commit, then:
git tag v0.2.1
git push --tags

The release.yml workflow runs tests, verifies the tag matches package.json, and publishes with signed provenance.

Node.js 18+ is required. Zero dependencies.

Full product documentation: https://cohesivity.ai/llms.txt