Claude Code
Cursor
Copilot
Codex
Windsurf
OpenCode
AntigravityLocal API clients
Choose a local API client when the main job is manual request exploration, desktop UX, and request collections.
Executable API documentation · API testing for coding agents
Apache-2.0 · Open Source Local-first · Built in Rust
Executable Markdown
for API workflows
Reqbook turns api-docs/*.md into runnable Markdown API specs that live beside your code. Humans get reviewable API docs and a local workspace; coding agents get compact files they can read, update, validate, and run without rediscovering your backend every session.
$ npm install -g reqbook
live walkthrough 
The problem
"Your agent can read the code. But it still has to rediscover how your API works every session."
Specs drift
API docs that live outside your codebase quietly go stale. The contract and the implementation stop agreeing.
Cloud lock-in
Login walls, team plans, and forced sync add friction for open-source teams and solo devs alike.
Agent blind spot
Your coding agent can read files, but not hidden app state in a cloud API workspace.
What you get
One binary. Plain markdown.
Every interface you need.
Design, run, and ship API contracts without leaving your repo, from the terminal, the browser, or your AI agent.
Markdown-native
api-docs/*.md commit with your code and review in pull requests. No proprietary format.
CLI + browser
rqb exec for scripts and CI. rqb serve for interactive design. Two interfaces, zero extra installs.
Agent-native
Claude Code, Cursor, Copilot, Codex, Antigravity, Windsurf, and OpenCode skills built in. Agents read, run, and write specs from the same files.
Flow canvas
Chain endpoints, capture values from one response, inject them downstream. Saved as markdown.
Import anything
Bring specs in from OpenAPI, cURL, or an existing export, or run rqb import project . to scan your codebase.
No cloud required
Rust binary, runs offline. No account, no subscription, no telemetry. Apache-2.0, yours completely.
Who it's for
Built for people who keep
their tools close to the code.
Backend devs
Specs that ship with code
Keep API contracts next to the handlers they describe, and validate them before you merge.
Agent users
Let your agent drive
Claude Code, Cursor, Copilot, Codex, Antigravity, Windsurf, and OpenCode inspect, update, and run specs from files, with no GUI automation.
OSS maintainers
No signup, no lock-in
No workspace account, no proprietary collection format, no forced cloud sync for contributors.
CI teams
Contracts in the pipeline
Run rqb validate and rqb exec on every pull request to catch drift early.
How it works
From empty repo to running specs
in three commands.
1
Init
Scaffold api-docs/ right inside your project, with no wizard and no signup.
$ rqb init --name=my-api \ --dev-url=http://localhost:8080 --yes
2
Write or import
Add specs as Markdown, or pull from OpenAPI, cURL, or an existing export. Nothing is locked in.
$ rqb import openapi ./openapi.yaml # → api-docs/imported.md
3
Run anywhere
Execute in CI, design in the browser, or install skills for your AI agent.
$ rqb exec api-docs/ # CI $ rqb serve # browser $ rqb skills install --agent=claude-code
The differentiator
An API workspace built for agent loops
Your agent should not spend every task reading routers, guessing payloads, and rebuilding curl commands. Reqbook gives it a surgical, executable API layer in the repo: fetch the contract, run the endpoint, diagnose failures, commit the markdown.
validate contracts fetch surgical context execute endpoints diagnose failures generate missing specs debug failing flows
# auto-detects agents and registers MCP tools $ rqb skills install $ rqb install mcp # give Codex/Claude/Cursor only the contract it needs $ rqb context refunds.quote --mode surgical --brief \ --max-fields 12 --include variables,request,response,errors,rules,verify # if a run fails, return likely cause + next action $ rqb diagnose api-docs/apis/refunds/post-quote.md --output json # ✓ executable API context ready
GUI collections are useful for exploration. Reqbook keeps agent context, execution, diagnosis, and reviewable API changes in Markdown files your repo, CI, and pull requests already understand.
Agent efficiency
Stop making agents
rediscover your APIs.
Without a shared API workspace, agents scan routes, inspect controllers, rebuild request bodies, generate cURL commands, and parse raw terminal output again and again.
- Read the exact endpoint file, not the whole backend.
- Run and validate a spec with rqb exec and rqb validate.
- Chain multi-step calls into a rqb flow.
- Get compact structured results, not noisy shell output.
- Fetch surgical context with rqb context --brief --max-fields 12 before reading backend source.
- Diagnose failures with rqb diagnose before scanning the whole service layer.
44.2% fewer total tokens in the latest Codex benchmark
Measured Jun 7, 2026 with Codex CLI 0.135.0 on examples/agent-token-api.
Source-only discovery used 96,599 total tokens; Reqbook surgical context used 53,857.
47.0% fewer uncached tokens 1.79x total-token ratio 1 run per scenario
Local benchmark, not a universal guarantee. Re-run on your stack before quoting broadly.
Positioning
Not another API client.
Executable API documentation.
Reqbook is for teams that want runnable Markdown API specs, API testing for coding agents, and reviewable API workflows in one source of truth.
The goal is not to clone a desktop API client. The goal is to make the API docs already living near your code executable enough for humans, CI, and agents to trust.
repo source of truth api-docs/
apis/users/post-create-user.md apis/auth/post-login.md flows/signup-login-profile.md $ rqb validate api-docs/ $ rqb flow api-docs/flows/signup-login-profile.md
HTTP runners
Choose an HTTP runner when the main job is terse command-line tests, captures, assertions, and CI output.
Reqbook
Choose Reqbook when API documentation should live in the repo, run as tests, review cleanly in PRs, and give coding agents stable context.
How it compares
Choose by workflow,
not by feature checklist.
API clients, HTTP runners, API platforms, and Reqbook solve different API problems. Reqbook is strongest when executable API documentation, PR review, CI, and agent context need to be the same artifact.
| Workflow need | Reqbook | Local API clients | HTTP runners | API platforms |
|---|---|---|---|---|
| Primary job | Executable API docs | Desktop request building | HTTP test execution | Team API operations |
| Source of truth | Markdown in repo | Collection files | Plain-text test file | Hosted or Git-backed workspace |
| Best fit for agents | Specs, MCP, skills | Request examples | CLI test context | Platform integration |
| PR review signal | Readable Markdown contract | Collection diff | Test file diff | Setup dependent |
| CI workflow | validate, exec, flow | Tool-specific CLI | Strong CLI reports | Platform CLI |
| When to choose it | Docs must run | You want a local app | You want terse tests | You need broad platform tooling |
Comparison reflects each tool's primary, out-of-the-box workflow. All trademarks belong to their owners.
Field notes
Latest from the blog
Practical notes on markdown API testing, agent workflows, and local-first API tooling.
API docs as code: executable contracts in your repo
A practical guide to API docs as code using Markdown contracts that developers can review, coding agents can read, and CI can execute.
Read the API docs as code: executable contracts in your repo guide CI API testing with Markdown specs
How to run CI API testing with Markdown API specs, stable exit codes, JUnit reports, and secrets passed through RQB_* environment variables.
Read the CI API testing with Markdown specs guide Writing agent-ready API docs in VS Code
A practical VS Code workflow for authoring Reqbook specs with variable autocomplete, previews, and agent context before coding agents touch the API.
Read the Writing agent-ready API docs in VS Code guideInstall
Install Reqbook.
Install the rqb binary directly with the shell installer, or use npm, crates.io, and the VS Code Marketplace. Then run specs from the terminal, browser UI, or VS Code.
Direct binary install Published
$ curl -fsSL https://markapidown.net/install.sh | sh $ rqb init --yes $ rqb serve
Installs the latest macOS or Linux binary and verifies its SHA-256 checksum when available. No Node.js or Rust toolchain required.
Release channels Live
- Shell installer
curl -fsSL https://markapidown.net/install.sh | sh - npm package
npm install -g reqbook - crates.io package
cargo install reqbook - VS Code extension
code --install-extension reqbook.reqbook-vscode - Source checkout
cargo install --git https://github.com/ngoclinh93qt/ReqBook
FAQ
Questions, answered.
Because plain .md files version with your code, review in pull requests, and stay readable to both people and AI agents. No proprietary format, no hosted collection behind a login.
No. Reqbook runs locally as a single Rust binary. There's no hosted workspace, no login, and no telemetry.
Install the rqb CLI directly with curl -fsSL https://markapidown.net/install.sh | sh, or use npm install -g reqbook and cargo install reqbook. The VS Code extension is available as reqbook.reqbook-vscode.
Yes. Reqbook ships agent skills so coding agents can read, write, validate, and run your API specs directly from the same Markdown files. Run rqb skills install for your agent of choice.
Yes. Import from OpenAPI, cURL, or an existing export, or run rqb import project . to scan your codebase and generate the specs you're missing.
Yes. Run rqb validate api-docs/ to check every spec on each push, and rqb exec to run them inside your pipeline.