Follow the steps below for your AI tool of choice - the whole process takes under a minute.
This flow works with any MCP-capable client:
Claude.ai
1 — Open connector settings
On Pro and Max, go to Customize → Connectors. On Team and Enterprise, go to Organization settings → Connectors.
2 — Add a custom connector
Click + → Add custom connector. Name it Birch and paste the URL: https://mcp.bir.ch/mcp
3 — Connect and sign in
Click Connect and sign in with your Birch account.
ChatGPT:
1 — Turn on Developer mode
In ChatGPT Web, open Settings → Plugins and switch on Developer mode.
2 — Add a custom connector
Go to Plugins and create a new one. Name it Birch and paste the URL: https://mcp.bir.ch/mcp. Choose OAuth, click Create.
3 — Connect and sign in
Click Connect and sign in with your Birch account.
Claude Code
1 — Run one command
Add Birch as a remote MCP server. No API key or bearer token needed.
In terminal:
claude mcp add --transport http Birch <https://mcp.bir.ch/mcp
2 — Finish OAuth
In an interactive session, run /mcp, select Birch, and choose Authenticate.
Codex
1 — Run one command
Add Birch as a remote MCP server. No API key or bearer token needed.
In terminal:
codex mcp add birch --url <https://mcp.bir.ch/mcp> && codex mcp login birch
2 — Finish OAuth
The login command opens the Birch authorization flow. Sign in, approve access, then start a new Codex session.
Once connected, you can manage which AI apps have access to Birch MCP from Settings → Birch MCP in Birch - including revoking access at any time.
Once you're connected, head to Birch MCP: Overview & Use Cases to see what you can ask your agent.
Getting new tools
We add new Birch MCP tools regularly. If you connected Birch MCP before a release, your AI client may not show the new tools straight away. This is expected:
Click Refresh tools in your MCP client.
If the new tools still don’t appear, disconnect Birch MCP and connect it again. In Claude.ai, go to Connectors → Birch MCP → Disconnect, then Connect again, and grant write access when the consent screen asks for it. In other clients, remove Birch MCP and add it again using the steps above.
Current limitations
Birch MCP is in Beta, so a few things to keep in mind:
No performance metrics. MCP returns what your rules did, not what it did to ROAS. For performance, use Birch AI in-app.
get_rule_logcaps the number of matched entities returned per run, so very large rule runs get truncated.Stage tools cover Meta, TikTok, and Snapchat – Google isn't supported yet.
Running a rule on demand works only for active rules (draft and paused rules can’t be run), and each rule can be run once every 10 minutes.
Creating new rules from your AI agent isn’t available yet – that’s on the roadmap, and when it ships, any AI-created rule will start as a disabled draft requiring your explicit confirmation before going live. You can already edit the conditions, tasks, and filters of existing rules, change their ad accounts, run them right away, and duplicate a rule as a starting point. For now, natural-language rule creation is available directly inside the Birch app.
Need help?
If you're still having trouble connecting Birch MCP, our customer care team will be happy to help. Contact us via the in-app chat or at support@bir.ch.