Skip to main content
This walkthrough is the developer playbook for the co-mind.ai API. Follow the steps in order: you sign in, mint one Personal Access Token, then use that single token for every capability the platform exposes — discovery, chat, embeddings, knowledge bases with retrieval-augmented chat, web research, and translation. Every step shows what to send and the response you should see. When something looks different in your terminal, you know exactly which step to revisit.
Endpoints referenced under /v1/. Replace {BASE_URL} with your deployment URL (given to you by your administrator).

​
The playbook

1. Log in

Sign in with email + password to receive a session JWT, then confirm the workspace your session is scoped to.

2. Create a PAT

Use the session JWT to mint a long-lived Personal Access Token with the scopes you need. Every step from here uses the PAT.

3. Discover models

List every chat and embedding model available on your deployment.

4. Compute embeddings

Turn text into vectors with an embedding-class model.

5. Chat completions

OpenAI-compatible chat — stateless, single-request-single-response.

6. Knowledge bases

Create a KB, upload files, query for context, and chat with retrieval-augmented generation.

7. Research

Single-shot web search and multi-step research sessions.

8. Translation

Translate text between languages with automatic source detection.

9. Manage tokens

List, rotate, and revoke your Personal Access Tokens.

​
Before you start

  • A running co-mind.ai deployment you can reach (see Installation)
  • User credentials — email and password, or an SSO session
  • curl in a terminal (any HTTP client works)
  • Background reading (optional but useful): Authentication, API Tokens
Ready? Start with Log in.