Connect your AI agent
Let Claude Code, Codex, Cursor, or an agent you build create, test, and publish trails in your workspace.
Trails are your organization's approved recipes for doing one job task with an AI tool. You can write them in the admin console, or hand the job to an agent you already run. The agent brings its own model and pays for its own tokens; Joring gives it the same authoring operations the console uses, behind a personal API key.
Two ways in, same operations:
| Best for | |
|---|---|
| MCP server | Claude Code, Codex, Cursor, Gemini CLI, VS Code, and any MCP client. Tools appear in the agent; no code to write. |
| REST API | Agents you build on the Anthropic or OpenAI APIs, scripts, and CI. OpenAPI document at https://api.joring.ai/openapi.json. |
Hosted connectors (claude.ai, ChatGPT) sign in with OAuth instead of a key: see OAuth for hosted agents.
Create an API key
In Settings → API keys, mint a personal key with the scopes the agent needs. How keys work.
Point your agent at Joring
Paste the snippet the key dialog shows you: one line for Claude Code, a JSON block for Cursor, a TOML block for Codex.
Ask for a trail
"Read the Joring authoring guide, then write a trail for account executives handling a competitor objection in ChatGPT. Run the surfacing examples until they pass." The agent drafts, tests, and asks before publishing.
Who can use it
API access is included in the Team plan and above, and is turned on per workspace. Anyone who can author trails (the Owner, Admins, Co-owners, and people holding the Trail author role) can create keys for themselves. Admins, IT admins, and Security admins can also see and revoke every key in the workspace. The roles are described in Teams and roles.
What an agent can do
Everything the trail editor can: list and read trails, create drafts, save documents, test which prompts surface a trail, publish and unpublish, restore earlier versions, and manage who can see a restricted trail. It cannot manage members, billing, or policies, and it cannot create more keys.
Every write shows up in the audit log attributed to the key and the person who minted it.