Open-source ADK
Turn any SaaS into an AI agent in minutes.
4emet is an open-source, CLI-first Agent Development Kit. Point it at your OpenAPI spec, answer a few questions, and get a LangGraph agent that talks to your API. No custom agent code required.
Building agents shouldn't take weeks.
You have a SaaS with a solid API. You want an AI agent that can use it—for support, workflows, or automation. Doing that from scratch means parsing OpenAPI, wiring auth, building tools, and maintaining prompts. 4emet does that for you: one init flow, one config file, and a ready-to-run agent.
Three steps to your agent.
Init
Run foremet init in your repo. Enter your OpenAPI URL (or path), product name, base URL, and auth. Pick your LLM (OpenAI, Anthropic, Gemini, or local Ollama). 4emet writes adk.config.json and generates the agent.
Configure
Copy .env.example to .env and add your SaaS API key and LLM key. No secrets in config—only env var names.
Run
Run foremet run for an interactive chat, or use the generated agent/run.py for scripts and integrations.
From zero to a working agent in under five minutes.
Built for API-first products.
OpenAPI-native
Point at any OpenAPI spec (URL or file). Tools and schemas are generated automatically.
LLM choice
Use OpenAI, Anthropic, Google (Gemini), or local Ollama. One config, same workflow.
LangGraph under the hood
Generated agents use LangGraph; you get a real graph-based agent, not a thin wrapper.
CLI-first
Init, run, and refresh from the terminal. Fits into scripts, CI, and existing dev workflows.
Refresh, don't rewrite
When your API changes, run foremet refresh to re-fetch the spec and regenerate the agent.
Secrets stay out of config
Config stores only the name of the env var; keys live in .env or your environment.
Where 4emet fits.
Support & help
An agent that looks up accounts, checks status, and explains your product using your real API.
Internal tools
Chat or script against your API without building custom UIs.
Automation
Combine natural language and API calls for workflows and integrations.
Demos & POCs
Show "AI + your product" in minutes instead of weeks.
Frequently asked questions.
Do I need to write agent code?
No. 4emet generates the agent (tools, system prompt, and runner) from your OpenAPI spec and your answers in foremet init. You only configure and run.
Which LLMs are supported?
OpenAI, Anthropic, Google (Gemini), and Ollama (local). You choose in init and set the matching API key (or use Ollama with no key).
What's in the generated agent?
LangChain tools per OpenAPI operation, a generated system prompt, and a run script. The agent runs as an interactive chat or via the script.
Can I regenerate when my API changes?
Yes. Run foremet refresh to re-fetch the OpenAPI spec and regenerate the agent. Your config is preserved.
Are my API keys stored in config?
No. Only the name of the env var (e.g. SAAS_API_KEY) is stored. You set the actual key in .env or your environment.
Your API. Your LLM. One agent.
Install the open-source ADK, run init, and go from OpenAPI to a working AI agent in minutes.
pip install foremetGet ADK on PyPI