Give Cursor, Claude, OpenAI Agents, Windsurf, and AI systems instant access to your APIs, SDKs, and product documentation. Paste a docs URL and generate a production-ready MCP server in minutes.
Works with Cursor, Claude Desktop, VS Code, Windsurf, OpenAI Agents and every MCP-compatible tool.
<60s
docs → MCP
PIPELINE
1 URL
paste & go
INPUT
Hosted
no install
REMOTE
5+
MCP clients
EXPORTS
Built for the modern AI ecosystem
One MCP server, ready to plug into every major AI editor and agent runtime.
Plus any MCP-compatible AI tool via manual configuration.
Watch how doc2mcp turns the entire Stripe documentation into an AI-ready MCP server — without writing a line of integration code.
Stripe Documentation
docs.stripe.com · 1,284 pages
Stripe Documentation
docs.stripe.com · 1,284 pages
Doc2MCP crawls content
Auto-discovers APIs, guides, SDKs
Doc2MCP crawls content
Auto-discovers APIs, guides, SDKs
Processes APIs + guides
Structured chunks, schemas, examples
Processes APIs + guides
Structured chunks, schemas, examples
Generates MCP server
23 tools · 6 workflows · 1 endpoint
Generates MCP server
23 tools · 6 workflows · 1 endpoint
Connects to Cursor
One-click config · remote MCP
Connects to Cursor
One-click config · remote MCP
AI understands Stripe instantly
No hallucinations · current APIs
AI understands Stripe instantly
No hallucinations · current APIs
No prompt stuffing.
No manual tool creation.
No hallucinated integrations.
Five steps from a URL to a production-ready MCP server, plugged into your agent of choice.
https://docs.stripe.com https://docs.langchain.com # doc2mcp toggle ON in chat
From crawl to retrieval to deployment — the full stack between your docs and an AI agent.
Automatically understands documentation hierarchy and structure — Mintlify, Docusaurus, Swagger, GitBook, raw markdown.
Generate MCP servers without writing MCP code. Tools, workflows, auth, hosting — done.
$ doc2mcp generate ./stripe-docs
→ crawled 1,284 pages
→ structured 4,182 chunks
→ tools: 23 · workflows: 6
✓ mcp ready · hosted
Improve AI coding accuracy and reduce hallucinations with semantic, schema-aware retrieval.
Keep documentation synchronized automatically — your MCP server tracks the source.
Works across the modern MCP ecosystem — Cursor, Claude, VS Code, Windsurf, OpenAI Agents.
Instant MCP configuration generation. Copy a JSON snippet, paste into your editor, done.
{
"mcpServers": {
"stripe": {
"url": "https://…",
"headers": { "Authorization": "Bearer …" }
}
}
}The compound effect of giving every agent — internal or customer- facing — structured, current, AI-ready documentation.
Agents respond from your real docs — not stale training data — so generated code actually compiles and ships.
Schema-aware retrieval narrows answers to the right endpoint, the right version, the right example.
Stop hand-writing tool wrappers per product. Doc2MCP generates them from documentation, automatically.
Live sync keeps the MCP server tracking your source — without redeploys, manual diffs, or token rotation.
Replace prompt stuffing with structured, retrievable, AI-optimized context — usable across every MCP client.
From “read docs” to “write integration” in minutes — across Cursor, Claude, Windsurf, OpenAI Agents.
Convert internal runbooks, ADRs, RFCs, and product docs into a queryable knowledge layer for your agents.
Stop pasting docs into prompts and debugging hallucinations — your team’s time goes back to product.
From public APIs to private wikis — doc2mcp converts any documentation surface into a hosted MCP server your agents can actually use.
Every API endpoint becomes a semantic tool your customers' agents can call directly — no SDK, no glue code, no support tickets.
Ship your docs as an MCP server alongside your dashboard. Cursor, Claude, and Windsurf can now answer questions about your product with first-party context.
Notion, Confluence, Markdown wikis — pipe them into a private MCP your engineers' agents can query without exposing data to third parties.
Drop your GitHub docs URL. Every contributor's Cursor or Claude instantly understands your project's architecture, API surface, and conventions.
Build it by hand and you ship a one-off, brittle integration. Use Doc2MCP and you ship infrastructure — structured, current, and ready for every agent in your stack.
Manual MCP development
DIY · slow · brittle
Doc2MCP
Infrastructure · automated · current
Setup time
Maintenance
Documentation understanding
Updates
Retrieval layer
AI readiness
Deployment
Support
| Capability | doc2mcpAI-native docs-to-MCP, the right way. | ArcMCPHosted search + get_page over Arc docs. | mcp-forgeManaged pre-built MCP servers + OpenAPI CLI. | Mintlify MCPSearch/edit Mintlify-hosted pages only. | StainlessRaw MCP servers from your OpenAPI spec. |
|---|---|---|---|---|---|
Smart Tools Endpoints compressed into semantic, AI-friendly toolkits (not raw POST /users). | Yes | Partial | Partial | No | No |
Workflow AI Infers auth, payment, upload, webhook, and CRUD flows — and surfaces them as agents. | Yes | No | Partial | No | No |
Visual Graph React-Flow graph of endpoint relationships, auth flow, and tool dependencies. | Yes | No | No | No | No |
Live Playground Test generated tools, simulate workflows, inspect outputs in-browser before shipping. | Yes | No | No | No | No |
AI Understanding Score Docs quality score · workflow confidence · auth confidence · MCP quality score. | Yes | Partial | No | No | No |
Live Processing UI Streaming terminal: crawl → analyze → detect auth → compress → workflows → MCP. | Yes | No | No | No | No |
Multi-Docs Engine Auto-detects Mintlify, Docusaurus, Swagger/OpenAPI, Markdown, GitBook, Postman. | Yes | No | Partial | Partial | Partial |
ASI1-native Built on ASI1 end-to-end. No OpenAI/Anthropic lock-in for the index layer. | Yes | No | No | No | No |
Endpoints group into User Management, Billing, Webhooks. Tools are LLM-shaped, not REST-shaped.
Customer Support Agent · Billing Agent · GitHub Automation · CRM Workflow — generated from your docs.
Watch crawl, parse, detect auth, compress endpoints, build MCP — all streaming in real time.
Researched from each vendor's own site, GitHub, and docs in May 2026. No marketing fluff — just scope, strengths, and honest gaps.
Hosted MCP for the Arc Network documentation — search and get_page only.
pricing · Free, scoped to Arc docs only
Where it wins
Where it falls short
Managed catalog of pre-built MCP servers (WordPress, Supabase, etc.) plus a coming-soon OpenAPI generator.
pricing · From $9/mo per server · $25/mo all-access
Where it wins
Where it falls short
Two MCPs scoped to docs already hosted on Mintlify — one read, one write.
pricing · Included in Mintlify plans
Where it wins
Where it falls short
Generates SDKs and MCP servers from your OpenAPI spec — endpoint-shaped.
pricing · Custom — built around SDK plans
Where it wins
Where it falls short
Tools per project
8–25
vs raw endpoint count
Time to live MCP
< 60s
from URL paste to Cursor
Inference scopes
6+
auth · payment · upload · webhook · crud
Sources: ArcMCP docs · mcp-forge.org & lorenzosaraiva/mcpforge · Mintlify MCP docs · Stainless MCP portal · evaluated May 2026.
The same MCP layer your engineers love — hardened with the controls, audit trails, and deployment options enterprise teams require.
Deploy doc2mcp inside your own VPC. Your docs and your generated MCP servers never leave your perimeter.
Single-tenant infrastructure on AWS, GCP, or Azure with bring-your-own-key encryption for every artifact.
SSO via SAML / OIDC, RBAC for projects + workspaces, and short-lived MCP tokens scoped to a specific team.
Every MCP query is logged with actor, tool, and document context — exportable to Datadog, Splunk, or S3.
Organize projects by team, share MCP servers across squads, and bill at the org level — no per-seat surprises.
Bring your own crawler, embedder, or post-processing step via webhook — doc2mcp orchestrates the rest.
See which tools your agents actually call, which docs are hot, and where retrieval quality is degrading.
SOC 2-ready controls, signed MCP token rotation, prompt-injection guards, and per-project secret isolation.
SOC 2 in progress
GDPR-ready data handling
EU + US regions
AI agents are becoming the next application layer. MCP is emerging as the standard protocol for tool access. Millions of APIs and documentation sites are not MCP-ready — Doc2MCP bridges that gap automatically.
“Documentation is becoming the knowledge layer for AI agents.”
Cursor, Claude, and Windsurf now write the majority of new code at AI-native startups. The interface to software is no longer a UI — it's an agent.
of new code at AI-native teams ships via agents
Anthropic, OpenAI, and the broader ecosystem have converged on the Model Context Protocol as the canonical way agents discover and use tools.
servers listed in the official MCP Registry
Millions of API references, SDKs, and developer portals were written for humans. There is no hosted, semantic, agent-ready layer in front of them.
of public docs expose a production MCP
We turn any docs URL into a hosted, agent-ready MCP server in under a minute — no manual coding, no infrastructure, no per-vendor lock-in.
from URL paste to a Cursor-ready MCP
Stats reflect internal observations across AI-native teams · May 2026
Coming soon: a first-class command line for doc2mcp. Convert docs, regenerate MCPs, and wire them into Cursor, Claude, and Windsurf — all scriptable, all in CI.
npm i -g doc2mcp · soon
doc2mcp — zsh
Every plan ships the same remote MCP. Bigger plans crawl more pages and re-crawl more often.
Auto-detected from your location · switch any time
For tinkering, demos, and one-off experiments.
For solo devs shipping side-projects.
Charged once as $3.99 · USD
For builders shipping AI-native products.
Charged once as $14.99 · USD
For teams operating internal docs at scale.
Charged once as $39.99 · USD
All plans include unlimited MCP read calls from Cursor. Conversions are the crawl + analyze step that runs once per docs site. Cancel anytime.
Everything builders ask before shipping with doc2mcp.
The Model Context Protocol is an open standard from Anthropic that lets AI agents (Cursor, Claude Desktop, Windsurf, OpenAI Agents) discover and call tools over a uniform JSON-RPC interface. doc2mcp produces fully compliant MCP servers your agents can connect to with a URL + Bearer token.
Paste a documentation URL. We crawl the site (Mintlify, Docusaurus, Swagger, GitBook, plain HTML — all supported), preserve code blocks, run an ASI1 understanding pass that detects auth, workflows, and endpoint groups, and generate a hosted MCP server with semantic tools — typically in under 60 seconds.
Anything that speaks MCP — Cursor, Claude Desktop, VS Code, Windsurf, OpenAI Agents, Cline, Zed, and the broader MCP client ecosystem. We provide ready-to-paste config snippets for every major client in your project dashboard.
Yes. Pro and Team plans support private projects with authenticated crawling and per-MCP Bearer tokens. Enterprise customers can self-host doc2mcp inside their VPC with SSO, RBAC, and audit logging — your docs and tokens never leave your perimeter.
Free plans run on-demand refreshes. Pro plans auto-resync every 24h, Team plans every 6h, and Enterprise plans support webhook-triggered instant updates. We use content-hash diffing so only changed pages get re-crawled.
Yes. We offer fully self-hosted deployments on AWS, GCP, or Azure with bring-your-own-key encryption, SSO via SAML/OIDC, audit logs, and custom retention policies. Reach out via the contact form to start a security review.
Still curious? Talk to the founders →
Stop building MCP servers manually. Generate AI-ready infrastructure directly from documentation — and let every agent your team uses speak the same language.
Built for the people building the agentic web.
5 +
MCP clients supported
v0.1
Live in the MCP Registry
ASI1
AI engine under the hood