REST API
Call Blazing Agents from a trusted backend over HTTP.
Overview
Every Blazing Agents feature is available over HTTP under /v1 at
https://api.blazingagents.com. Use the REST API from a backend written in a
language without an SDK, or when you want direct control of headers and
streams. Set BLAZING_AGENTS_BASE_URL to that origin to run the cURL examples
on these pages.
Keep bearer credentials on the backend
A Blazing Agents API key grants access to your whole tenant. Never put it in browser or mobile code; route end-user requests through your backend.
curl "$BLAZING_AGENTS_BASE_URL/v1/agents" \
--header "Authorization: Bearer $BLAZING_AGENTS_API_KEY"Conventions
Requests and responses use JSON, except avatar uploads, which use multipart form data. Generation and session turns stream their output. Every request is scoped to the tenant that owns the credential.
Every endpoint in this reference accepts a ba_ API key. The pages are
generated from the platform's OpenAPI contract, which is also served at
/openapi.json. Endpoints that only the dashboard can call, such as API key
management, are not listed here.
Every response carries an X-Request-Id header in req_… format. Log it with
your application logs and include it when you contact support. Each HTTP
attempt gets a new value, so a retry has a different ID, and you cannot set it
yourself.
To attach your own correlation ID, send X-Client-Request-Id with 1–128
letters, digits, ., _, :, or -. Blazing Agents logs it next to the
request ID. It is not an idempotency key.
Errors use the standard error envelope. A stream that has already started reports later failures inside the stream.
Resources
| Resource | Operations |
|---|---|
| Agents | Configuration, versions, lifecycle, avatars, and MCP attachments |
| Workspaces | Workspace lifecycle and network policy |
| Skills | Skill creation, archives, files, and copies |
| Tenant | Current identity and settings |
| Providers | Model provider keys and model discovery |
| MCP connections | Connection lifecycle, testing, and OAuth |
| Generation | Stateless text and structured output |
| Sessions | Turns, transcripts, and tool approvals |
| Artifacts | Listing, download, and deletion |
| Memories | Notes an agent keeps between sessions |
| Prompts | Reusable message templates |
| Usage | Tenant and agent usage queries |
| Monetization | Merchant connection, customer bindings, and usage events |
| Tasks | Background task definitions and schedules |
| Task runs | Execution, polling, and cancellation |
| Chat connections | Slack and Telegram connections and delivery outcomes |
Next
- Authentication to choose a credential.
- Errors to handle failures.
- Streaming to read turn output.
- TypeScript SDK or Python SDK to skip raw HTTP.