Playbooks
Find a playbook by what it does, build one from a description, check its data flow before it runs, and publish it as the live version.
create_playbookverify_playbookpublish_playbookrun_playbook +6 more
Running ops automations at
Reach your whole Ballet workspace by asking
The connector exposes 58 tools. Claude picks the ones it needs, so you describe the outcome rather than naming them.
Runs
Read a run back step by step: what each one returned and where it stopped.
list_runsget_runget_run_steps
Data tables
The tables your playbooks keep, from the accounts they've handled to a queue of work waiting its turn. Read them, add to them, or set up a new one.
list_data_tablescreate_data_tablelist_data_table_recordsupdate_data_table_record +9 more
Reports
Open the reports in your workspace and ask what they found, without opening a dashboard.
get_reportlist_reports
Agents
The agents a playbook calls on, each with its own instructions, tools and model.
create_agentupdate_agentlist_agentslist_models +2 more
Connections
Connect the systems a playbook works across, through Ballet's own integrations or any MCP server, with keys kept in the workspace vault and never read back.
list_integrationsconnect_mcp_servercreate_mcp_servercreate_secret +7 more
Tools, skills and guides
The building blocks a playbook composes, and Ballet's own building guides, which Claude reads first so what it builds follows how Ballet works.
create_toolcreate_skillget_builder_guiderun_scratch_code +9 more
Come back later and just ask it to run
Once a playbook exists, a conversation is one of the ways to start it. Ask by what it does — Claude finds it, runs it, and reads the result back.
Ask for it the way you'd ask a colleague
You don't need the playbook's name or its id. Claude searches what your workspace has and picks the one that fits.
The answer comes back in plain language
Not a run id to go and look up. What ran, what it did, and the one thing that needs a person.
It's still a governed run
The gate held because the workflow says it should. Asking from a conversation doesn't route around it.
Ask what a report found, and what's still waiting
The connector reads more than playbooks. Ask about your workspace's reports or the tables your playbooks keep, and the answer comes back in a sentence.
Build it in a conversation, then let it run without one
A playbook built from Claude is a playbook like any other. It doesn't need a prompt to start, and it doesn't need Claude to be open.
- A schedule
- A webhook
- A Slack command
- A form submission
- Manually, from Claude
The same reviewed playbook
What Claude runs is the version your engineers approved, and the gates they set still hold.
Your workspace, your sign-in
The connector authenticates against your own Ballet workspace, so Claude sees what you see.
Every run still logged
A run started from a conversation is logged and replayable like any other.
Pick the door your team already uses
The playbooks and the governance are the same in all of them.
Platform
Build and run the workflow
Where a workflow gets written, reviewed, and watched once it's live.
Platform →
Chrome extension
Run playbooks beside the tab you're in
A side panel next to your CRM or helpdesk. Ask for something, or type / and pick a playbook.
Chrome extension →
Claude connector
Build and run playbooks inside Claude
Connect Ballet in Claude and ask it to run a playbook, read what happened, or build a new one.
You're here
Slack agent
Tag Ballet where the work already happens
Mention Ballet on a message, or send it a DM. The result posts back in the thread.
Slack agent →
Add Ballet to Claude
The connector is in the directory. Adding it takes a sign-in, and Claude can reach your playbooks from the next message.