Skip to main content

Honeybase documentation

Honeybase is a workflow-automation platform: you design processes, release them as workflows, and every execution is a run — with human tasks, teams, schedules and availability on one side, and AI agents on the other.

Get started​

Three ways in, depending on who you are. Pick your path — you can always add the others later.

Use the appCreate an account

Sign up or sign in at app.honeybase.ai to build processes, run workflows and manage your organization in the browser. Start here if you are new.

Drive it with AIConnect the MCP

Let an AI client drive Honeybase directly — list processes, read runs, create tasks — over the Model Context Protocol, with OAuth handling the sign-in.

Write code against itGet an API key

Create an organization API key (hbk_…) and call the GraphQL API from your own code. One endpoint, one key, an explicit permission set.

Explore the docs​

These pages cover the ideas the product is built on, what people build with it, the systems you can connect it to, and every operation the API exposes.

Start hereConcepts

Processes, workflows, runs, tasks, agents — what each one is, how they fit together, and the vocabulary the rest of these docs assumes.

See it appliedUse cases

Worked examples of automations teams actually run, from the trigger to the humans and agents in the middle to what comes out the other end.

Connect your systemsIntegrations

Databases, Slack, WhatsApp, email, Google Sheets, any HTTP API, and the LLM provider behind the AI features. One page per type, listing exactly what the setup form asks for and which credentials to obtain first.

Build against itGraphQL API

One endpoint, one organization API key. Authentication, conventions, pagination and errors, runnable examples, and a generated reference for every type, query and mutation.

Two things worth knowing up front​

Credentials never live in your workflow definitions. Secrets go to a dedicated secret store when you create the connection; workflow nodes reference the connection, not the credential. Reading a connection back through the API returns its configuration with secret values masked.

Everything an API key does is attributed to that key. Task writes made with a key show the key's name in the task's audit trail — you do not impersonate a user, and you cannot.