canonical: https://jentic.com/apis/autumn.ai/autumn # Autumn Task API The Autumn Task API runs durable research tasks carried out by a sandboxed agent. A caller starts a task, lists tasks, reads a task's status, state, and outputs, and continues, re-executes, or stops it. Streaming routes deliver task events as server-sent events, a credits endpoint reports the remaining balance, and upload endpoints presign and validate files for a task. Errors share one body shape so a caller can branch on the error and retry when the response marks it retryable. ## For AI agents Run and track durable research tasks executed by a sandboxed agent: start tasks, stream their events, read status, state, and outputs, and continue, re-execute, or stop them. Authenticate with an X-API-Key header or a bearer key. ## Scope Does not expose a raw chat completion endpoint, train models, or store files long term. Use for running and tracking durable agent research tasks only. ## Capabilities - Start a durable research task run by a sandboxed agent - Stream task events as the agent works through server-sent events - List tasks and read a task's status, state, and named outputs - Continue or re-execute a task with new input - Stop a running task - Read the account's remaining credits - Presign a file upload and validate it for a task ## Use cases ### Run a Durable Research Task An agent starts a research task on Autumn, streams its events as the sandboxed agent works, and reads the output when it finishes. Because tasks are durable, the agent can start work, disconnect, and poll the status later rather than holding a long-lived request open, which suits research that takes time to complete. Example prompt: Call POST /task to start a research task, then GET `/task/{task_id}` to poll its status and GET `/task/{task_id}/output` to read the result ### Stream and Continue Long Tasks An agent streams a task's events in real time and, when a result needs refining, continues the same task with additional input instead of starting over. Autumn keeps the task state, so the agent builds on prior work and reads intermediate state as the task progresses. Example prompt: Call POST `/task/stream` to start and watch a task, then POST `/task/{task_id}/continue` to extend it with new input ### AI Agent Integration via Jentic An AI agent reaches the Autumn Task API through Jentic without holding the key itself. It searches by intent, receives the matching operation with its input schema, and executes the call with the credential injected at run time, so starting and tracking research tasks becomes a step in a larger agent workflow. Example prompt: Search Jentic for 'start a research task', load the POST /task schema, and execute it with Jentic-managed credentials ## Key endpoints | Method | Path | Description | | --- | --- | --- | | POST | `/task` | Start a task | | GET | `/task` | List tasks | | POST | `/task/stream` | Start a task and stream its events | | GET | `/task/{task_id}` | Get task status | | POST | `/task/{task_id}/continue` | Continue a task | | POST | `/task/{task_id}/stop` | Stop a task | | GET | `/task/{task_id}/output` | Get task output | | GET | `/credits` | Get remaining credits | ## Key resources - **Tasks** — Start, list, continue, execute, and stop tasks via /task, `/task/{task_id}/continue`, and `/task/{task_id}/stop` - **Task State and Output** — Read a task's status, state, and outputs via `/task/{task_id}`, `/task/{task_id}/state`, and `/task/{task_id}/output` - **Credits and Uploads** — Read credits and prepare files via /credits, /upload-url, and /validate-upload ## AI readiness This API is usable in Jentic One now. Its AI-readiness score against Jentic's framework shows where it stands today and where improvements would make it even easier for agents to use. - **Score:** 70 / 100 - **Maturity:** AI-Aware - **Dimensions:** - Foundational Compliance: 99 / 100 - Developer Experience & Jentic Compatibility: 74 / 100 - AI-Readiness & Agent Experience: 55 / 100 - Agent Usability: 94 / 100 - Security: 57 / 100 - AI Discoverability: 61 / 100 - **View full report:** https://jentic.com/apis/autumn.ai/autumn/scorecard - **How the score is calculated:** https://docs.jentic.com/reference/api-readiness-framework/overview/ - **More about the dimensions:** https://docs.jentic.com/reference/api-readiness-framework/specification/#dimensional-model-overview ### Score it yourself Every API in the directory is allowlisted, so you can re-score it with no key required. - **Score your own API:** https://jentic.com/scorecard.md - **Scoring CLI agent skill:** https://github.com/jentic/jentic-api-scorecard/blob/main/skills/jentic-api-scorecard/SKILL.md ```sh npx @jentic/api-scorecard-cli score ``` ## Why Jentic - **Setup:** Wiring the Autumn Task API by hand means choosing between its X-API-Key and bearer schemes, handling server-sent event streams, and polling durable tasks for completion yourself. Through Jentic you install once, import Autumn from the API Directory, store the key once, and your agent calls it. - **Permission scoping:** Autumn puts the task id in the URL path, so a rule can pin your agent to the task operations you choose. You choose the operations it may call, such as starting a task and reading its status and output, so continue, re-execute, and stop are not included unless you add them. - **Credential handling:** Your Autumn key is stored once, encrypted, by your own Jentic One instance and injected at execution time. It never enters the agent's prompt, logs, or context. - **Discovery method:** Agents search Jentic by intent such as 'start a research task' or 'read a task's output', and Jentic returns the matching Autumn operation with its input schema so the agent calls the right endpoint without browsing the reference docs. ## Related APIs - **Perplexity** — AI answer engine for grounded research questions - **OpenAI** — Language models for chat, reasoning, and tool use - **Anthropic** — Claude models for chat, reasoning, and tool use ## FAQ ### What authentication does the Autumn Task API use? The Autumn Task API authenticates with an API key in the X-API-Key header or a bearer key in the Authorization header, per its OpenAPI spec, and a few operations also accept a session token. Through Jentic the key is stored encrypted by your own Jentic One instance and injected at execution time, so it never reaches the agent. ### Can I start and stream a research task with the Autumn Task API? Yes. POST /task starts a durable task and POST `/task/stream` starts one and streams its events as server-sent events. You can then read status with GET `/task/{task_id}` and the result with GET `/task/{task_id}/output`, or continue the task with more input. ### What are the rate limits for the Autumn Task API? Autumn meters work against a credit balance that GET /credits reports rather than a fixed request ceiling, and the OpenAPI spec does not state request limits. Read your credits before a batch, watch for HTTP 429 responses, and check Autumn's documentation for the current limits. ### How do I start a research task with the Autumn Task API through Jentic? Search Jentic for 'start a research task', and it returns the Autumn operation with its input schema. Load the schema and execute POST /task with the credential injected at run time. To run it on your own infrastructure, install Jentic One from its GitHub repo. ### Is there an Autumn Task API MCP server? You don't need an MCP server to give your agent the Autumn Task API. Jentic connects it directly from the API Directory: import it, store your key once, and your agent calls the task operations it needs. Operations are discovered on demand, so no extra server's tool definitions sit in the agent's context. ### Can I limit what my agent is allowed to do with the Autumn Task API? Yes. Autumn puts the task id in the URL path, so a rule can allow only starting tasks and the read operations that track them, while continue, re-execute, and stop stay off the table until you add them. Every call runs through your own instance where it is logged for review.