Integrations · Scripts, CI and other agents

No browser? Use a key.

Claude connects through a connector you approve in the browser. A CI job, a script or another vendor's cloud agent cannot do that, so it uses an API key instead: the same MCP endpoint, the same tools, under its own name.

Read only or read and writeA daily write capRevoke any time

How the sync works

Two places, one truth.

01

One endpoint

Every agent talks to https://api.projectify.si/mcp over MCP (JSON-RPC over HTTPS). A key goes in the Authorization header as a Bearer token.

02

Its own name

Each key is its own agent on the Agents page, acting for the person who made it. You see its calls, its writes today and when it was last used.

03

Same rules

Claims, Always lines and activity logging work exactly as they do for Claude. A key never bypasses a rule.

Set it up

Three steps.

New to Projectify? Getting started walks through the whole setup, with pictures.

1

Create a key

In Projectify, open Agents › Connect Claude › Create a key instead. Pick Script or CI, choose read only or read and write, and copy the key. It is shown once.

2

Store it as a secret

Put the key in your platform's secrets (a CI secret, an environment variable), never in the repo. If the platform limits outbound hosts, allow api.projectify.si.

3

Call the tools

Point any MCP client at https://api.projectify.si/mcp with the header Authorization: Bearer <key>, or post JSON-RPC to it directly: tools/list, then tools/call with start_session.

What your agents see

A key-connected agent gets the same tools and briefing as Claude. MCP is plain JSON-RPC over HTTPS, so a shell script with curl is enough.

Using Claude Code on the web? You do not need a key: the connector on your Claude account already reaches cloud sessions.

Questions

Scripts, CI and other agents and Projectify, answered.

More in the full FAQ.

Can a script, CI job or another cloud agent connect?

Yes, with an API key. Create one on Agents › Connect Claude › Create a key instead, store it as a secret and send it as a Bearer token to the MCP endpoint. How it works.

Does it work with assistants other than Claude?

Projectify speaks MCP, so any MCP client can connect. Claude Code, claude.ai and Claude Desktop are the ones we build and test with today.

Can an agent see projects I have not given it?

No. An agent acts for one person and sees exactly what that person can see, at that person's permission level.

One key. The same board.

Free for one person. Revoke a key any time.