Does AI know how to use your product?API?
Redocly scores your API description for agent-readiness, flags what an agent would get wrong, and publishes it the way agents and developers read it: a reference with try-it, an MCP server, SDKs and workflows tested against production. All from the API description you already have, with every seat included.
- Reference + try‑ittry-it for OpenAPI, GraphQL
- Mock serverlive before the backend is
- SDKsEarly accessTypeScript, Python, Go, PHP
- Workflows + testsArazzo, run against production
- Lint + scoreagent-readiness, on every PR
- MCP serveryour docs, as tools
Change the description and all of these change with it.
How it works
Find the gaps before an agent does.
An agent only knows what your API description tells it. Redocly scores the description on every pull request and flags the parts an agent can get wrong. It gives agents your API in the forms they use: MCP tools, agent skills and SDKs (early access). Workflows run against production prove the docs match the API.
Proof
Agents are already reading API docs.
Measured across docs sites published with Redocly over the last 30 days.
- 9 in 10
- docs sites get traffic from AI agents and crawlers every week
- of sites with human traffic
- 7 in 10
- docs sites served coding agents through their MCP server
- last 30 days
- 1 in 4
- requests to the most visited docs sites now come from AI, and 1 in 5 on a typical site
- top tenth of sites
- 93%
- of questions to the AI assistant answered from the docs, 99% of those answers with a citation
- last 30 days
Per-site ratios: a small site counts as much as a large one.
What you publish
Your customers integrate faster.
Write Markdown, Markdoc or React pages next to your API description in git. Redocly builds the site, previews every pull request, and keeps the reference and the mock server in step with the description.
Developers make their first successful call without leaving the page
Reference for OpenAPI, GraphQL and AsyncAPI, with try-it for OpenAPI and GraphQL
One site, one search, one navigation. Readers never land in a second product
Markdown, Markdoc and React pages next to the reference
Your customers integrate before your backend ships
Mock server generated from the description
Your API in your customer's language, published the day the spec changes
TypeScript, Python, Go and PHP SDKs
You find out the API changed before a customer does
Arazzo workflows run against the mock and production
Every team ships to the same standard, and you can see who doesn't
Catalog with scores and owners
Two kinds of AI
One AI for the people who use your API. Another for the people who maintain it.
Both are in beta and included with every plan at no extra cost, within fair-use limits. Capabilities that need Scale or Enterprise are marked.
Readers get answers from your docs, with the source cited
AI Assistant beta · in-docs answers with citations, embeddable, MCP server
- In your docs, every plan
- Embeddable widget for your app, Scale and up
- MCP server so agents read docs as tools, every plan
How do I refund part of a payment?
Use POST /refunds with amount below the original. Partial refunds need the payment to be captured.
Every pull request is checked against the API before a human reads it
Reviewer beta · reviews pull requests, drafts fixes, sets up new projects
- Pull request reviews, every plan
- Drafted fixes as pull requests, Enterprise
- New project setup, every plan
Production returned "brewing" in 3 of 40 test runs. The description was missing a state. Fix drafted.
Keep it current
Docs that never lie.
Your docs live in git next to the API description, and every push is previewed, checked and tested.
Review docs the way you review code
Preview deployment per branch
Docs from many repositories, published as one site
Remote content from other repositories, URLs and CI pipelinesScale
Docs stay true because they are tested, not because someone remembered to update them
Workflows flag where the API and the description disagree
Catch a regression before it merges
Style rules and the agent-readiness score on every pull request
For the organization
Every team on one standard.
Roll it out without a procurement cycle: Scale is self-serve, paid by card, with single sign-on, a catalog, scorecards, multi-product sites and analytics. Enterprise adds role-based access, SCIM, audit logs, data residency and an annual contract.
Private docs your whole company can open, and nobody else can
SAML and OIDC single sign-on for readers and editors
Every API has an owner and a score
Catalog and scorecards against your standards
See what readers look for, and what the assistant could not answer
Search, page and assistant analytics
Control who sees and edits every page
RBAC, SCIM provisioning and audit logsEnterprise
Pricing
Pay for the product, not the people.
Every plan has unlimited seats, unlimited readers and unlimited git contributors. Inviting a colleague is never a billing event. The invoice has your plan and the extra projects you added, and nothing else.
- Editors and admins
- Readers, public or signed in
- Git contributors and previews
Pro
For the team shipping an API and the docs that go with it. 3 projects, unlimited pages.
What Pro includesMost popular
Scale
For the organization rolling docs out across teams. SSO, catalog, multi-product sites, early-access SDKs. 10 projects.
What Scale includes
Enterprise
Access control, compliance, data residency and support. Annual contract.
What Enterprise includesAPIs documented with Redocly
Ernie ParkEngineering Manager, Brex"The main thing is that our API reference is extremely easy to update. Engineers just worry about the code. Right now I can keep the docs up to date with very minimal work."