Integration · Direct your agents
Codex builds what you chose, not what it guessed.
Add Annsa to Codex and it starts from the priority you picked, with the evidence behind it. Same list, same spec as Claude Code and Cursor.
What it does for you
Switch agents without re-explaining.
Choose what to build.
Codex starts from the priority you picked, with the quotes, people and findings behind it.
Direct your agents.
You or Codex write the spec from the evidence. Codex reads your decisions and what you ruled out, and asks you when a call is yours.
Learn as you go.
It files what it finds back into the same list as your customers and your other agents. A person makes the call.
The proof
Connect, then ask.
Add the lines, run the login, and ask for your priorities. The first reply tells you which workspace it is talking to.
[mcp_servers.annsa] url = "https://app.annsa.ai/mcp" auth = "oauth"How it connects
A few lines in config.toml.
Add the server, sign in once, then ask for your priorities.
Add the server.
In `~/.codex/config.toml` add `[mcp_servers.annsa]` with `url = "https://app.annsa.ai/mcp"` and `auth = "oauth"`.
Log in.
Run `codex mcp login annsa` and approve in the browser. Codex signs in as you.
Check it worked.
Ask Codex for your priorities. The reply’s `who` should say production, app.annsa.ai and your workspace.
Try this first.
“Show me the top priority and why it ranks there.” Then: “I’ve chosen #N. Read its spec and build it.”