I built an open-source tool that turns your OpenAPI-documented REST API into a conversational interface

Project Showcase
1

Most applications already expose the actions users need through REST APIs.

But users still have to navigate dashboards, find the correct screen, fill out forms, and perform several clicks to complete a task.

I built Agentify , an open-source tool that adds a conversational interface on top of an existing REST API.

GitHub: GitHub - feature1-ai/agentify-by-f1: Turn your OpenAPI/Swagger spec into a conversational agent for your SPA — no code, human-approved actions. · GitHub

The setup is:

  1. Add your OpenAPI or Swagger specifications to the project.
  2. Configure your REST API base URL and authentication.
  3. Deploy Agentify.
  4. Connect the included chat interface or integrate it into your own SPA.

Users can then describe what they want in natural language.

For example:

Cancel order 1023.

Agentify reads the OpenAPI specification, identifies the relevant endpoint and parameters, and shows the proposed API call to the user.

Nothing is executed immediately. Every action pauses for human approval before Agentify calls the API.

Agentify does not connect directly to your database. It operates entirely through the application’s existing REST API, so the API’s authentication, authorization, validation, rate limits, and audit rules continue to apply.

The project currently includes:

  • OpenAPI and Swagger-based API discovery
  • Natural-language-to-API mapping
  • Human approval before execution
  • A built-in chat interface
  • Per-user or server-level API credentials
  • REST, webhook, and SSE interfaces
  • Docker-based deployment
  • LangGraph workflows
  • Codex CLI as the current reasoning runtime
  • A framework-agnostic client for React, Vue, Svelte, or vanilla JavaScript

I think this could be useful for developers building conversational interfaces for:

  • SaaS products
  • Internal tools
  • Admin dashboards
  • API-first platforms
  • Applications with complex workflows

I’d appreciate feedback from developers working with OpenAPI, AI agents, REST APIs, LangGraph, or human-in-the-loop systems.

In particular:

  • Would you integrate this into an existing application?
  • Is requiring approval before every API action the right default?
  • What security concerns would you want addressed?
  • Which agent runtime should be supported next?
  • What application would you try it with?

Contributions and architecture feedback are welcome.