← All help topics

MCP bridge and AI agents

Connect your AI agent (e.g. Claude Desktop) to TestIP over MCP and manage posts right from your chat.

1🧩 How it works

TestIP itself is not an MCP server: it gives AI agents a plain REST API (routes /api/agent/...). So that an MCP-capable agent can use it, a ready, lightweight MCP bridge is included (the mcp-server folder).

The chain looks like this:

  • 🤖 your AI agent (MCP client)
  • 🔗 MCP bridge — you run it locally
  • 🟢 TestIP — over the network, with your token

The bridge exposes the tools: schedule_post, find_post, approve_post, reject_post.

You don't need to host a separate «server with a URL» — the bridge runs locally and talks to TestIP itself.

2🔒 Step 1. Create a token

  1. Open «Project info».
  2. Section «API tokens for AI agents».
  3. Click «Create token».

⚠️ The token is shown once — copy it right away.

3⚙️ Step 2. Run the MCP bridge locally

In the mcp-server folder run npm install, then start the bridge with two environment variables:

  • TESTIP_API_BASE — your TestIP service URL, e.g. https://testip.example
  • TESTIP_TOKEN — the token from step 1

The bridge talks to the agent over stdio (standard input/output), so it needs no separate URL.

4🔗 Step 3. Connect your MCP client

In your MCP client config (e.g. Claude Desktop) add a server:

  • command — node
  • argument — the path to mcp-server/index.mjs
  • the same env vars: TESTIP_API_BASE, TESTIP_TOKEN

✅ After restarting the client, TestIP tools become available to the agent.

5🔒 Security

  • Only the token hash is stored; the full token is shown once.
  • A token can be revoked at any time.
  • The agent's actions are limited to the token's project scope.

Didn't find an answer? Write to us — we'll help.

Contact support