Skip to content

Docs overview

Cuttlely is the open-source agent platform for applications. It gives your applications their own intelligent capabilities and coordinated agent teams. A team is a lead agent, Harness Prime, that splits each question among specialists, and in Cuttlely it is called a harness. Your app calls a team through the API, and people can chat with it on a share link or a website embed. Knowledge pipelines (ETL), visual workflows (chatflows and agentflows) and Build with AI give the teams what they need, and every change is a draft until you publish it. It all runs on infrastructure you control.

New here? Start with Your first 10 minutes, or open the bakery team in about a minute with Try the demo. Building an app? Start with the API guide, then Use a team from your app. The harness Prediction API section is the reference for harness responses and stream events.

  • Harnesses. A team of specialists that Harness Prime sends work to. Build one from a description, try it, and call it from a chatflow.
  • Analyst mode. A specialist with its own database that creates tables and answers from them. Turn it on with CUTTLELY_ANALYST=auto; on Render, see Analyst database.
  • ETL. Pipelines that load, clean, split, embed and store your data for search, on a schedule if you like.
  • Canvas cards. Chatflows and agentflows, compact or classic cards (CANVAS_CARD_LOOK=classic sets classic for everyone), Build with AI, and older V1 agentflows.
  • Version history. Drafts and Publish, the check before Publish, restore and roll back, a copy in GitHub, and Dev to Prod.
  • Share and embed a chat. A public link for any chatflow, a chat button or full-page chat on your website, and how the share settings change the look.
  • Use the API in your own web app. API keys, calling chatflows, agentflows and harnesses, streaming, conversation memory, uploads, errors and allowed origins, with a runnable example app.
  • Docker. The install most people want. SQLite, one container.
  • From source. For changing Cuttlely.
  • Render. A hosted install with Postgres.
  • Server settings. Change how the server runs from Admin > Server settings: sign-in and AuthKit, the database, storage, the analyst and more, with a review step, safe boot and History.
  • Sign in. Turn on the local account, create the first account in a terminal or the browser, and sign in. Set up WorkOS AuthKit when a team signs in.