Will it slow down my API?
No. The SDK sends logs in the background, and fix lookups happen in-process, with no network call in the request path.
For teams with a public API on Node, Python, Ruby, or Go
Catch every failed agent call and put your fix right in the error response. Agents correct the call and finish the job, so nobody reports your API as down.
It logs every call, shows you which ones fail, and lets you answer the agent right in the error response. It works alongside the docs you already have on ReadMe, Mintlify, Fern, or GitBook.
Your error messages were written for developers reading logs later. Agents read them the instant a call fails and act on whatever they say. A bare 400 gives them nothing.
It hits a wall: The agent calls an endpoint that doesn't exist or sends a field your API rejects, and gets back invalid_request.
It guesses: With no hint in the error, it retries the same bad call, tries something else, or gives up.
You never see it: The failure never reaches your team, so you can't tell what the agent was trying to do.
The ticket says your API is down. Most agents start each session fresh, so the next one makes the same bad call. The person who asked the agent for help leaves with nothing done, and you never hear why.
| What you have | What it does | Where it stops |
|---|---|---|
| llms.txt and Markdown docs | Helps agents read your docs | The call still fails |
| A hosted MCP server | Gives agents your endpoints as tools | Returns the same bare error |
| Your error tracker | Records the 400 for your team | Tells the agent nothing |
| Better error text in code | Fixes one message | Every change waits on a deploy |
Restless works alongside ReadMe, Mintlify, Fern, and GitBook. Keep your docs where they are.
See the failed calls. Every request lands in Restless, so you can see which calls fail and where callers give up.
Write the fix once. Plain language, with no code change and no redeploy.
Every agent gets it. Restless adds your fix to the error response for every matching call. The agent corrects the request and retries.
Nothing reaches an agent until you approve it.
The task finishes, so nobody writes in to say your API is down.
Every call in one place See which endpoints agents hit, which calls fail, and exactly where callers get stuck along the way.
Agent feedback Agents flag missing endpoints and confusing docs to an inbox that syncs to Linear.
Agent-ready docs Restless builds agent-readable docs and a hosted MCP server from your OpenAPI spec.
Sign up via
Google or GitHub
Paste a prompt into your coding agent or CLI
Install Restless and onboard your APIs
No. The SDK sends logs in the background, and fix lookups happen in-process, with no network call in the request path.
Your API keeps serving traffic, because the SDK fails open. And we never drop your traffic for billing reasons.
The SDK strips personal data before anything leaves your server, and the SDK is open source. Logs are kept for 30 days.
You do. Nothing reaches an agent until you approve it.
Free plan
We're a small team, so only the first 100 teams get in. Once they're in, the channel closes.
If agents already handle your API fine, the free plan will show you that too. If they don't, you'll see the failed calls as soon as traffic arrives, and you can fix the first one the same day.