One URL, zero tokens: OAuth comes to JITM.ai
Connecting an AI agent to JITM.ai now takes exactly one step: paste the URL. OAuth handles the rest. No tokens to generate, no config files to edit, no secrets to rotate.
Until today, every agent that wanted to use JITM.ai needed a Personal Access Token. You'd log into the dashboard, generate a token, copy it, paste it into your agent's config, and hope you didn't accidentally commit it to a repo. It worked, but it was one more thing standing between "I want to try this" and actually trying it.
That barrier is gone. The MCP server at mcp.jitm.ai now supports OAuth. Paste the URL into Claude Desktop, Cursor, ChatGPT, Windsurf, or any MCP client that supports OAuth, and you'll be prompted to sign in with your JITM.ai account. That's it. The agent is connected and authenticated. All 13 tools, ready to go.
Why this matters for security
Tokens are static secrets. They sit in config files, environment variables, and clipboard histories. If one leaks, it works until someone notices and revokes it. OAuth replaces that with short-lived sessions tied to your identity. The agent never sees a long-lived secret because there isn't one. Sessions expire, consent can be revoked, and your account's security policies (MFA, session timeouts) apply automatically. This is the foundation that lets us build features like per-agent audit trails and scoped permissions down the road.
Why this matters for agents
The fewer steps between an agent and a tool, the more likely the agent actually uses it. With token-based auth, an agent needed the user to leave the conversation, visit a dashboard, generate a token, and come back. With OAuth, the agent asks to connect, the user approves in a browser popup, and the conversation continues. The agent can go from "I've never heard of JITM.ai" to "your model is trained" without the user ever leaving the chat.
Tokens still work
If you're running agents in CI pipelines, scripts, or headless environments where there's no browser to sign in with, Personal Access Tokens are still fully supported. Pass them via ?token=pat_xxx or the JITM_MCP_TOKEN environment variable, same as before. The token path is now the fallback rather than the default, but nothing about it has changed. Existing integrations keep working.
Try it now
Open your MCP client, add https://mcp.jitm.ai as a server, and sign in. No dashboard visit needed. Your agent is ten seconds away from training its first model.