Agentify

Agent Interface Compiler — One command. Every agent speaks your product.

Agentify compiles any OpenAPI specification into 8 agent interface formats — MCP Server, CLAUDE.md, AGENTS.md, .cursorrules, Skills, llms.txt, GEMINI.md, and A2A Card. Instead of hand-building each format separately, generate them all from a single source of truth.

npx agentify-cli transform https://petstore.swagger.io/v2/swagger.json

The Problem

AI agents are the new users of your API. But making your product agent-accessible requires building and maintaining multiple interface formats:

That's 8+ formats to build, test, and keep in sync. Every API change means updating all of them.

The Solution

Agentify is a compiler. OpenAPI in, every agent format out.

                    +---> MCP Server (with Dockerfile)
                    |
                    +---> CLAUDE.md
                    |
                    +---> AGENTS.md
                    |
OpenAPI Spec -----> +---> .cursorrules
                    |
                    +---> Skills
                    |
                    +---> llms.txt
                    |
                    +---> GEMINI.md
                    |
                    +---> A2A Card

Quick Start

# Transform any OpenAPI spec
npx agentify-cli transform https://petstore.swagger.io/v2/swagger.json

# Specify output directory
npx agentify-cli transform ./my-api.yaml -o ./output

# Override project name
npx agentify-cli transform https://api.example.com/openapi.json -n my-project

# Generate only specific formats
npx agentify-cli transform ./my-api.yaml -f mcp claude.md

Output:

  Agentify v0.3.0
  Agent Interface Compiler

  +-- 20 endpoints detected -> SMALL API strategy
  +-- 3 domains identified (pet, store, user)
  +-- Auth: apiKey (SWAGGER_PETSTORE_API_KEY)
  +-- Strategy: Direct tool mapping — one tool per endpoint

  > Generated mcp + claude.md + agents.md + cursorrules + llms.txt + gemini.md + skills + a2a (15 files)
  > Output: ./swagger-petstore-mcp-server
  > Security scan: PASSED

Features

Smart Strategy Selection — Automatically chooses the right generation strategy based on API size:

Security First — Every generated artifact passes through:

• Input sanitization (blocks eval, exec, Function constructor, require/import injection)
• Handlebars template injection prevention
• Prompt injection pattern detection
• Generated code security scanning

Production Ready — Generated MCP servers include:

• TypeScript source with full type safety
• Dockerfile for containerized deployment
• Environment variable configuration (.env.example)
• Stdio transport (standard MCP protocol)

Output Format Status

How It Works

1. PARSE        OpenAPI 3.x / Swagger 2.0 spec (URL or file)
                  |
2. SANITIZE     Strip dangerous patterns from all spec fields
                  |
3. ANALYZE      Detect domains, auth, scale -> pick strategy
                  |
4. COMPILE      Generate AgentifyIR (intermediate representation)
                  |
5. EMIT         Run selected emitters (MCP, Skills, Docs, etc.)
                  |
6. SCAN         Security scan all generated code
                  |
7. OUTPUT       Write files to disk

AgentifyIR is the canonical intermediate representation — a flat, typed structure that captures everything an emitter needs: product metadata, capabilities (endpoints), domains, auth config, and generation strategy.

Architecture

agentify/
+-- src/
|   +-- cli.ts              # CLI entry point (Commander.js)
|   +-- parser/             # OpenAPI parsing + input sanitization
|   +-- generator/          # Pluggable emitters for each format
|   |   +-- templates/      # Handlebars templates
|   +-- security/           # Input sanitization + output scanning
|   +-- types.ts            # AgentifyIR type definitions
+-- templates/              # Generated project templates
+-- test/                   # Vitest test suite

Contributing

Agentify welcomes contributions, especially new emitters (output formats). Each emitter implements a simple interface:

import type { Emitter, AgentifyIR, EmitterOptions, EmitterResult } from "../types";

export class MyFormatEmitter implements Emitter {
  readonly name = "my-format";
  readonly format = "my-format";

  async emit(ir: AgentifyIR, options: EmitterOptions): Promise<EmitterResult> {
    // Generate output files from the IR
    return { filesWritten: [...], warnings: [] };
  }
}

See CONTRIBUTING.md for the full guide.

Roadmap

• M0: Foundation — OpenAPI parser, MCP emitter, security scanner, CLI
• M1: Multi-Format — CLAUDE.md, AGENTS.md, Skills, .cursorrules, llms.txt, GEMINI.md, A2A Card
• M2: Intelligence — Capability graph, semantic grouping, context optimization
• M3: Self-Serve — Web UI, one-click deploy, registry integrations
• M4: Scale — Enterprise features, custom emitters, CI/CD integration

Compared to Alternatives

No existing tool compiles one OpenAPI spec into all agent interface formats.

License

MIT -- Agentify Contributors