Coding Agents

What it is

You create a personal access token in Settings → Developer. The editor sends that token with every tool call. Setod answers the call and does not run a model of its own — the editor's model does the thinking.

Install the Setod skill as well. It tells the editor the order to follow. Live details (what each connector tool returns, plan limits, failure patterns) come from the server, so an installed skill does not go stale.

Connect Claude Code

Copy the Claude Code command from Settings → Developer. It looks like this:

claude mcp add --transport http setod <url> --header "Authorization: Bearer <token>"

1. Create a token and copy the command. The token is shown only once.
2. Run the command in a terminal, then open Claude Code in the project.
3. Install the skill: /plugin marketplace add mamipour/setod-skills, then /plugin install setod-fde@setod.

Connect Cursor

Settings → Developer has a ~/.cursor/mcp.json block. Paste it into that file (or into the project's .cursor/mcp.json), with your token in the Authorization header. Restart Cursor so it picks up the server.

Install the skill with npx skills add mamipour/setod-skills. Later, npx skills update pulls a newer skill when the server says yours is behind.

Install the skill

• Cursor and most other clients: npx skills add mamipour/setod-skills
• Claude Code: /plugin marketplace add mamipour/setod-skills, then /plugin install setod-fde@setod

What it can and cannot do

It can list and edit agents, attach connectors you already connected, write prompt skills and code skills, dry-run an agent, and publish after a passing dry run. It can read runs and explain why one failed.

• It cannot create connectors, invite members, change billing, or delete agents.
• It cannot read or write code-skill secret values.
• A live run (one that really sends mail or messages) only happens if you ask for it.
• Changing tools on an already published agent takes effect on the next run. The editor must ask you first.

One token is one workspace

A token belongs to you and to one workspace. If you work in a second workspace, add a second server entry with its own token. The editor cannot switch workspaces with the same token.

Tokens and revoking

Tokens expire after 30, 90, or 365 days, or never. You can hold ten active tokens per workspace. Revoke one from Settings → Developer; the editor's next call fails until you paste a new token. Signing out of the dashboard also invalidates them (a sign-out resets your session everywhere), so create a fresh token after you log back in.

The plaintext is shown once, in the browser tab that created it. Setod stores only a hash.