Adaptora - Open Source
Connect any API. Talk to it in plain English.
9 followers
Connect any API. Talk to it in plain English.
9 followers
Self-hosted Dynamic API Agent with an MCP layer — talk to any REST API in natural language. Bring-your-own-LLM (Ollama). - guptayush02/adaptora
How does Adaptora handle auth flows for APIs that require OAuth2 instead of just a static API key, and does that work out of the box with Ollama or do I need to set something up myself?
@ecrin_t68691 Great question! Adaptora extracts OAuth2 URLs (authorize + token) from the API docs automatically. The PKCE flow is implemented on the backend — browser redirect, callback, token storage, auto-refresh.
That said, I'm still testing the full OAuth2 client credentials flow end-to-end. LinkedIn currently works with a Bearer token you paste in directly. Full "click Connect → browser login → auto token" is built but needs more real-world testing before I can say it works out of the box for every provider.
If you're an early adopter and hit issues, open a GitHub issue — I'm actively fixing these. 🙏
Love that this is self-hosted and Ollama-friendly, exactly what I've been hunting for. One thing that would make it stickier for me: built-in webhook support so my adapted endpoints can push events out to n8n or other automations, not just respond to queries.
@emirdemirbwv23
Thanks, glad it's hitting the spot! Webhook support is a great call — right now adaptora is pull/request-based only, but pushing events out to n8n/Zapier-style automations is exactly the kind of thing that'd make it feel like a real integration hub instead of just an adapter layer. Adding it to the roadmap. If you've got a specific use case (which events, what payload shape you'd want), drop it in a GitHub issue — helps me design it around real workflows instead of guessing.
Does it handle auth flows like OAuth2 automatically, or do I need to set up tokens myself for each API?
@gencalmasa49194 Right now you set up the token yourself — paste your Bearer token or API key in the Connect screen, and Adaptora handles every call after that automatically (auth headers, token refresh, retries).
Full OAuth2 "click Connect → browser login → auto token" flow is built in the code but I'm still validating it across providers before calling it production-ready.
Tried it with Ollama and a small local model, and it actually picked up my endpoints without much handholding. The MCP layer is a nice touch, makes spinning up new API connections feel quick.
@asel843239 This is huge feedback 🙏 Ollama + small model working smoothly is exactly what we were aiming for — no need to throw GPT-4 at every API call. And yeah, the MCP layer is the game-changer; it keeps everything plugged and swappable.
More people testing this locally is what we need. Thanks for giving it a real shot!