A second opinion on your agent's work.
Write down what good looks like, in plain language. Judge reads a piece of finished work against those criteria and returns a verdict, so a result is checked before anyone relies on it.
Judge
What an agent does with it
-
Takes your criteria
Acceptance criteria written the way you would brief a person.
-
Reads the work
Checks a finished deliverable against each one.
-
Returns a verdict
A structured answer an automation can act on: pass, fail, or not sure.
Judge
Questions people ask about judge
When would I use this?
The other built-in apps
-
To-dos
A shared working list on boards. Agents plan, start, block and complete to-dos, and every change lands in an activity log.
-
Sales CRM
Organizations, contacts, opportunities and every interaction, with the evidence behind each claim kept beside it.
-
Memory
What an agent learns stays learned: state, facts and notes it can search and come back to.
-
Web research
Search the web, read a page, find similar ones, and run deeper research when a question needs it.
-
Webhooks
A delivery URL any system can post to. Each delivery is captured and can start an automation.
-
Contact enrichment
Find and validate a work email, and look up similar companies from a domain.
-
Database query
Ask a question of a Postgres database you connect, in plain language, and get a chart back.
More of the platform
Each capability has a page. These are the ones nearest this.
-
Workflow automation
Scheduled agents that run a described process
-
MCP apps
Apps your team builds, delivered over MCP
-
Built-in apps
To-dos, a CRM, memory and more, ready for agents
-
Skills
Business judgment as version-controlled files
-
Integrations
Connectors for the systems you already run
-
MCP gateways
Remote MCP servers from any registry
-
Security
Workspace isolation and per-user access