OpenAI-Compatible API
TAIBLES exposes an OpenAI-compatible HTTP API. Point any OpenAI SDK or client at your TAIBLES base URL and authenticate with an API key.
This is not another model router. The same API surface reaches:
- Your assistants and agents — including the default assistant with direct tool use, knowledge access, and automation-building abilities, or any agent you build in TAIBLES
- Plain chat models — when you pass a model URL (for example
openai://…), the request runs through the default assistant with that model for the turn
Compatibility is the transport. Agents are the product.
Base URL
https://<your-taibles-host>/api/openai/v1On taibles.ai, that is typically https://taibles.ai/api/openai/v1. Self-hosted and private deployments use your own origin. The in-app Settings → API Keys page shows the exact base URL for your environment.
Authentication
Create an API key under Settings → API Keys. Keys use the sk-tbl- prefix. Send them as a Bearer token:
Authorization: Bearer sk-tbl-<your-key>API keys act on your behalf: your agents, memories, files, and organization context are available to the request.
Endpoints
| Method | Path | Description |
|---|---|---|
GET | /v1/models | List agents and chat models available to your key |
POST | /v1/chat/completions | Create a chat completion (JSON or SSE when stream: true) |
What model means
The model field selects what runs:
model value | Behavior |
|---|---|
default-assistant | Your organization's default assistant |
Agent slug (e.g. support-bot) | That agent (slug from the agent name) |
| Agent UUID | That agent by id |
Chat model URL (e.g. openai://gpt-5.4) | Default assistant with that LLM for the turn |
Use GET /models to see ids for your org.
Next steps
- Quickstart — create a key and call
default-assistant - Agents and models — how model ids resolve
- Chat completions — streaming, tools, TAIBLES extensions
- Examples — SDK
base_urldrop-ins