# API to Agents

> API to Agents is a managed service that audits a company's API and turns it into an MCP (Model Context Protocol) server that ChatGPT, Claude, and other AI agents can use safely. We design, build, host, secure, and maintain the server under the customer's own subdomain, such as mcp.customer.com. API to Agents is a brand of Neoval GmbH, Inselstrasse 49, 4057 Basel, Switzerland.

## What we offer

- Free Agent Readiness Audit: a fully automated, two-minute assessment that scores an API out of 100 across API quality, authentication, tools, and testing. It proposes the MCP tools an assistant would get (read, write, and guarded tools with confirmation boundaries), lists risks and blockers with mitigations, and gives a roadmap with a delivery estimate.
- Fixed setup quote: calculated by versioned, deterministic pricing rules (never by a model), itemized, valid for 14 days, from $499. The base scope covers one upstream API, up to three read tools, API-key authentication, and one revision round. Additional write, sensitive, or OAuth tools are priced by the same rules.
- Managed hosting: Starter $29.99 per month and Growth $99.99 per month, covering hosting, monitoring, maintenance, and updates on EU infrastructure. Billing starts only when the integration is live. Enterprise self-hosting or private cloud is scoped on request.
- After payment, a human-validated integration roadmap is published to the customer dashboard within 72 hours.

## How it works

1. Verify a work email, answer four short questions, and share API documentation (public URL or upload). No API keys or credentials are requested.
2. Receive the readiness score, proposed MCP tools, risks, roadmap, delivery estimate, and setup quote automatically.
3. Pay the setup fee through Stripe Checkout. Our team validates the scope and publishes the integration roadmap within 72 hours.

## Who it is for

Companies with a documented REST or GraphQL API (webhooks, SDKs, and private integration APIs can qualify) that want their service reachable from ChatGPT, Claude, and other MCP clients without building MCP expertise in-house.

## Pages

- [Home](https://apitoagents.com/): product overview, how it works, pricing
- [Free audit](https://app.apitoagents.com/audit): start the Agent Readiness Audit
- [Website builder](https://apitoagents.com/website-builder): Coming soon: build a complete website from ChatGPT or Claude, with frontend, backend, storage and deployment, and own the source code to extend, move or keep maintaining from the chat.
- [REA reverse engineering](https://apitoagents.com/rea-mcp): Explore REA, the original MCP toolkit for authorized software analysis, our fork, and local setup paths for ChatGPT, Codex, and Claude.
- [Dedicated REA MCP](https://apitoagents.com/rea-mcp/hosted): Get your dedicated REA MCP endpoint: connect ChatGPT, Claude and other remote clients to static software analysis without installing REA or Ghidra.
- [MCP authentication](https://apitoagents.com/mcp-authentication): Connect gated content to ChatGPT and Claude with existing accounts, clear consent, and permission-aware MCP tools.
- [Managed MCP server](https://apitoagents.com/managed-mcp-server): Hosting, security, monitoring, and maintenance for the MCP server that connects your API to AI agents, under your own subdomain.
- [API to MCP](https://apitoagents.com/api-to-mcp): How a REST, GraphQL, or OpenAPI-documented API becomes a set of MCP tools, and what is automated versus designed.
- [Agent Readiness Audit](https://apitoagents.com/agent-readiness-audit): The free two-minute audit: what it scores, what it delivers, and how it leads to a fixed setup quote.
- [ChatGPT integration](https://apitoagents.com/integrations/chatgpt): What it takes for ChatGPT to search, read, and act on your service through an MCP server.
- [Claude integration](https://apitoagents.com/integrations/claude): How Claude connects to a remote MCP server and what your customers can do with your service inside Claude.
- [Pricing](https://apitoagents.com/pricing): Every setup line item and hosting plan, with the versioned rules that produce each quote.
- [For online shops](https://apitoagents.com/ecommerce): What a shop gets when customers can search, compare, add to cart and order from ChatGPT or Claude, and what it costs to get there.
- [Preview your tools](https://apitoagents.com/preview): Paste an OpenAPI, Swagger or Postman URL and see the tools an assistant would get, classified as read, write or guarded, in seconds and without signing up.
- [Agent-ready checklist](https://apitoagents.com/agent-ready): Twelve verifiable checks an API and its MCP server must pass before assistants can use it safely, with a free self-declared badge.
- [Built in public](https://apitoagents.com/built-in-public): The evidence behind the service: four tested MCP servers in public, a conformance scenario run in CI, measured generator benchmarks, and published prices.
- [MCP glossary](https://apitoagents.com/mcp-glossary): Plain definitions of MCP server, tool, client, guarded tool, agent readiness, managed hosting, and the other terms used across this site.
- [Security overview](https://apitoagents.com/security): no customer credentials in the audit, safe document ingestion, verified access and payment, EU infrastructure
- [Contact](https://apitoagents.com/contact): Enterprise scoping and company information
- [Blog](https://apitoagents.com/blog): practical guides on MCP tools, OAuth, testing, and agent-ready API design
- [Terms](https://apitoagents.com/terms) and [Privacy](https://apitoagents.com/privacy)

## Documentation

- [Documentation index](https://apitoagents.com/docs): build an MCP server for your API in TypeScript, Python, C# or PHP; tested examples at https://github.com/apitoagents/mcp-examples
- [What an MCP server is, and what yours has to do](https://apitoagents.com/docs/what-is-an-mcp-server): Learn what an MCP server exposes, how assistants call its tools, and the responsibilities that turn an API integration into a reliable service.
- [Tool design: reads, writes and guarded actions](https://apitoagents.com/docs/tool-design-reads-writes-guarded): The three kinds of tool every integration needs, the rules each one follows, and the four-tool interface that all the example servers implement identically.
- [The Sample Shop API every example wraps](https://apitoagents.com/docs/sample-shop-api): Explore the Sample Shop API used by our MCP examples: product search, order lookup, an idempotent cart update, and a guarded checkout action.
- [Build an MCP server in TypeScript](https://apitoagents.com/docs/build-mcp-server-typescript): Build a TypeScript MCP server for the Sample Shop API with typed tools, annotations, error handling, and a Streamable HTTP endpoint using Express.
- [Build an MCP server in Python](https://apitoagents.com/docs/build-mcp-server-python): Build a Python MCP server for the Sample Shop API with typed tools, input descriptions, actionable errors, and a Streamable HTTP endpoint.
- [Build an MCP server in C# (.NET)](https://apitoagents.com/docs/build-mcp-server-csharp): Build a C# MCP server for the Sample Shop API with annotated tools, parameter descriptions, error handling, and an ASP.NET Core HTTP endpoint.
- [Build an MCP server in PHP](https://apitoagents.com/docs/build-mcp-server-php): Build a PHP MCP server for the Sample Shop API with attribute-based tools, input schemas, actionable errors, and a PSR-7 HTTP entry point.
- [Conformance testing: one scenario, four servers](https://apitoagents.com/docs/conformance-testing): Test four MCP servers with one shared scenario. Check tool schemas, annotations, errors, and behaviour across TypeScript, Python, C#, and PHP.
- [Connect your MCP server to Claude](https://apitoagents.com/docs/connect-claude): Connect a remote MCP server to Claude on the web, desktop, or Claude Code. Test with the MCP Inspector and check the workflow before launch.
- [Connect your MCP server to ChatGPT](https://apitoagents.com/docs/connect-chatgpt): Prepare your remote MCP server for ChatGPT. Review HTTPS access, authorization, connection setup, and the tool behaviours to test before launch.
- [OpenAPI-to-MCP generators compared, on one API](https://apitoagents.com/docs/openapi-to-mcp-generators-compared): Compare two OpenAPI-to-MCP generators on the Sample Shop API: tool names, schemas, annotations, authentication, errors, and runnable output.
- [Go to production](https://apitoagents.com/docs/go-to-production): Prepare your MCP server for production with HTTPS, authorization, permission checks, secret storage, monitoring, scaling, and client compatibility.
- [How REA works](https://apitoagents.com/docs/rea-how-it-works): A plain explanation of REA, the open-source reverse engineering MCP server: the investigation loop, what your assistant receives, and how evidence stays honest.
- [REA's major components](https://apitoagents.com/docs/rea-components): The parts of REA: the CLI, the stdio MCP server, the analysis providers it wraps, agent setup and skill, readiness checks, and the evidence it keeps.
- [Requirements to run REA locally and in the cloud](https://apitoagents.com/docs/rea-requirements): What REA needs on a laptop: Node.js, engines per target, memory and OS notes. And what changes when you host it: transport, isolation, licences and sizing.
- [What REA can and cannot do](https://apitoagents.com/docs/rea-possibilities-limits): Realistic expectations for REA: the questions it answers well per target, what it will not do, and the limits set by engines, formats and the law.
- [Website builder overview](https://apitoagents.com/docs/website-builder-overview): What the website builder is, what a conversation with ChatGPT or Claude produces, who owns the result, and what is in the coming-soon release.
- [Connect from ChatGPT or Claude](https://apitoagents.com/docs/website-builder-connect): How the website builder appears in ChatGPT and Claude as a remote MCP connector, the tools it exposes, and what a first conversation looks like.
- [Your code and repository](https://apitoagents.com/docs/website-builder-your-code): What is inside a generated website repository, how the builder and your developers work in it at the same time, and the rules that keep its history clean.
- [Transfer and maintain](https://apitoagents.com/docs/website-builder-transfer-and-maintain): How to move a generated website to your own infrastructure, and how to keep maintaining it from your chat afterwards or without the builder at all.

## Articles

- [Hosted REA and Website Builder MCP for ChatGPT and Claude](https://apitoagents.com/blog/hosted-rea-website-builder-mcp): Explore hosted REA MCP for authorized software analysis and our coming-soon website builder for ChatGPT and Claude, from a brief to a reviewed preview.
- [REA: Reverse Engineering with ChatGPT, Claude, and MCP](https://apitoagents.com/blog/rea-mcp-reverse-engineering-ai-assistants): Learn how REA connects software analysis to ChatGPT, Codex, and Claude through local MCP, with an authorized Electron example and practical setup steps.
- [Self-hosted vs managed MCP server: a decision guide for product teams](https://apitoagents.com/blog/self-hosted-vs-managed-mcp-server): Compare self-hosted and managed MCP servers across ownership, security, costs, and maintenance. Use six questions to choose the right setup for your team.
- [MCP for SaaS: what it costs, who it is for, and what you get](https://apitoagents.com/blog/mcp-for-saas-what-it-costs): Understand MCP integration costs for SaaS: setup, authentication, documentation, hosting, and maintenance, with published prices and worked examples.
- [How to connect your API to ChatGPT](https://apitoagents.com/blog/how-to-connect-your-api-to-chatgpt): Connect your API to ChatGPT using GPT Actions or an MCP server. Compare the options, add an optional app interface, and test a complete customer workflow.
- [MCP server hosting: self-hosted, platform, or managed?](https://apitoagents.com/blog/mcp-server-hosting-options-compared): Compare self-hosted, platform, and managed MCP hosting. Review costs, security, maintenance, and who owns the endpoint after your first demo.
- [Add AI Project Status Reports to Your SaaS with MCP](https://apitoagents.com/blog/saas-project-status-reports-mcp): Turn project tasks, deadlines, and blockers into useful AI status reports. Design an MCP workflow with clear scope, source links, and permission checks.
- [Make Your Ecommerce Shop Available in ChatGPT and Claude](https://apitoagents.com/blog/make-your-ecommerce-shop-available-in-chatgpt): Let customers search your inventory, compare products, add to cart, and check out. Learn how an MCP server brings your ecommerce shop into ChatGPT and Claude.
- [MCP vs REST API: Where Each Fits in Your SaaS Architecture](https://apitoagents.com/blog/mcp-vs-rest-api-service-architecture): Learn when to keep a direct API integration, when an MCP adapter helps assistants, and how to share business rules without creating a second backend.
- [MCP Tool Pagination: Help Assistants Handle Large Result Sets](https://apitoagents.com/blog/mcp-tool-pagination-large-result-sets): Design paginated MCP tools with continuation tokens, bounded results, permission checks, and honest answers when only part of a dataset is available.
- [Testing an MCP integration: build a realistic evaluation set](https://apitoagents.com/blog/testing-mcp-integration-evaluation-set): Build an MCP evaluation set around user goals, synthetic fixtures, allowed actions, observable outcomes, and repeatable tests of ambiguity and failure.
- [MCP tool errors: distinguish empty results, invalid input, and unavailable services](https://apitoagents.com/blog/mcp-tool-errors-empty-results-retries): Design MCP tool responses that tell assistants when to adjust inputs, report no matches, wait, or stop, with concrete examples and a practical test checklist.
- [Idempotency for agent actions: design retries without duplicate bookings](https://apitoagents.com/blog/idempotency-agent-actions-duplicate-bookings): Prevent duplicate bookings and orders with idempotent MCP tools. Design stable operation IDs, payload checks, concurrent retries, and recovery.
- [MCP OAuth: authentication, authorization, and content access](https://apitoagents.com/blog/oauth-remote-mcp-connection-record-access): Learn how MCP OAuth connects ChatGPT and Claude to private content, with clear consent, token validation, scopes, and record-level permission checks.
- [Read-only MCP tools still need permissions: a practical design guide](https://apitoagents.com/blog/read-only-mcp-tools-permissions): Design read-only MCP tools with record access checks, limited output fields, scoped search, and concrete tests for cross-account data exposure.
- [How to write MCP tool descriptions that help assistants choose correctly](https://apitoagents.com/blog/writing-mcp-tool-descriptions): Write clearer MCP tool descriptions with a before-and-after example, a reusable template, precise input guidance, and a practical evaluation checklist.
- [OpenAPI to MCP: what can be automated, and what still needs design](https://apitoagents.com/blog/openapi-to-mcp-what-can-be-automated): Learn what an OpenAPI-to-MCP converter can reuse, where mappings need review, and how to validate a generated integration with a concrete search example.
- [How to choose which API actions to expose as MCP tools](https://apitoagents.com/blog/choosing-api-actions-for-mcp-tools): Choose your first MCP tools with a practical workflow: map customer requests, separate reads from writes, define results, and test failure cases.
- [From navigating to asking: how chat is quietly replacing the way people use the web](https://apitoagents.com/blog/from-navigating-to-asking-how-chat-changes-habits): Explore how conversational assistants change the way people use the web, and what service owners need to make their products usable through chat.
- [From plugins to MCP: how agents are changing the way people use your service](https://apitoagents.com/blog/how-mcp-servers-apps-and-connectors-change-services): Understand how MCP servers, apps, and connectors let assistants use your service, and what this means for teams that own APIs and customer workflows.

## Full content

- [llms-full.txt](https://apitoagents.com/llms-full.txt): this file plus the complete text of every article