Agent builder
Describe the job in a sentence. It runs without you.
A saved AI worker with its own tools, schedule, and channels.
Read the agent docs- One sentence
- Drafts the configuration
- Toolboxes
- Scoped per agent
- Cron or event
- How a run triggers
- Admin-gated
- Only admins create agents
Clado
Olive
Codebuff
Attensira
Assemble
Opennote
Stratify
Gatekeep- Clado
- Olive
- Codebuff
- Attensira
- Assemble
- Opennote
- Stratify
- Gatekeep
- Clado
- Olive
- Codebuff
- Attensira
- Assemble
- Opennote
- Stratify
- Gatekeep
Prompt-first builder
The agent that builds agents.
Describe the job and the builder streams a draft into the form: which toolboxes it needs, what it should be told, where the output goes. Or start from a template and change what you disagree with.
- Draft the whole configuration from one sentence
- Templates for the jobs teams ask for most
- Test the agent before you save it
Toolboxes
An agent is only as useful as what it can reach.
Skills read your project data with no OAuth. Apps need a project connection. Channels decide how the result reaches a person.
Product Analytics
Query events, sessions, users, and funnels — the same data behind the dashboard, not a summary of it.
Watch Sessions
Watch session replays for friction, confusion, and bugs, and cite the moment it found them.
Issues
Triage runtime, repo, behavioral, and visual issues with severity, affected users, and a reproducing replay.
Coding Agent
Writes a fix and opens a pull request through the Human Behavior GitHub App — into a branch, never your default.
Agent Builder
Build, configure, and invoke other agents from natural language, so one agent can stand up the next.
Your connected apps
Linear, Jira, GitHub, Notion, Stripe, HubSpot, Supabase and the rest of the connector catalog, scoped per agent.
How it runs
Configuration, then unattended work.
- 01
Give it tools
Attach only the toolboxes the job needs. Enabling a connector on the project does not give every agent access to it.
- 02
Give it a trigger
A cron schedule in your timezone, an event trigger from your product, or nothing at all if you would rather just chat with it.
- 03
Give it an audience
Slack, email, SMS, or WhatsApp. Findings land where your team already is instead of in a dashboard nobody opens.
Details
How agents behave in a real team.
- Runs are conversations
- Every scheduled run lands in the agent's Recent chats as a thread you can open, read, and continue — including the tool calls it made.
- Memory is opt-in
- Turn on memory and the agent keeps directives and files across runs. Leave it off and each run starts clean.
- Admin-gated
- Only project admins can create agents or set agent instructions, and only owners can connect integrations.
- Upcoming runs are visible
- An unpaused schedule pins its next occurrence to the agent page, so a silent agent is obviously paused rather than mysteriously quiet.
Keep going
The rest of the platform.
Everything here reads from the same recorded session — there is nothing to stitch together between surfaces.
Stop guessing what happened.
One SDK, one afternoon. We will walk your team through it on a call and leave you with the recording of your own product.