Pro Features

Every account runs the fast loop: agents are briefed, work, commit outcomes, and confidence moves. The paid tiers add the machinery that makes the loop improve itself — grading, repair, safe change, training — and the controls a team needs to run it at scale. This page is the map. Each row says what the capability does for the loop and where to find it.

Making the loop recursive

Capability	What it adds	Where
Behavior	Judges grade runs, incidents open on recurring failures, and fixes to the learning are proposed, replayed against the failing traces, shipped and rolled back. Pro proposes fixes; Teams and Enterprise can ship them automatically after replay or a canary.	Agent page → Behavior
Learning Branching	Branches, pull requests, tags and rollback over what agents learn, so a change can be isolated, reviewed and undone.	Agent page → Git Repository
Learning Consolidation	Demotes superseded and stale learning automatically and proposes merges of convergent knowledge for review.	Git Repository → Branches → Consolidation
Managed Models	A private model fine-tuned on the decisions that worked, scored against your own history, served behind one model string. Pro: one live model; Teams: three.	Models
ML Layer	Learned ranking, per-account calibration of how much outcomes count, and training-data export.	Pro Tools

Seeing what the loop knows

Capability	What it adds	Where
Immutable Decision Traces	Every run sealed, signed and chained, with LLM calls, cost and OpenTelemetry export.	Traces, and Runs on the agent page
Knowledge Graph and Expertise	Paths and queries across the whole graph, and which agent knows what.	Knowledge Graph, Expertise Map
Pattern Detection	Conflicts, stale knowledge, single points of knowledge and recurring failures across your agents.	Patterns
Intelligence Layer	Entity extraction, the memory critic, the distiller, pre-write validation and multi-strategy retrieval.	Pro Tools, Entity Extraction
Learnings Health and Learning Tiers	The confidence, composition and integrity of everything learned, and which of it is hot, warm or archived.	Learnings Health, Learning Tiers

Running it as a team

Capability	What it adds	Where
Accounts and Access	Account isolation, roles, teams, scoped API keys, rate limits and the audit log.	Settings
Learning Security	Secrets, personal data, poisoning and contamination found in what agents learned, with quarantine and revert.	Security
Rooms	Shared workspaces where several people’s agents learn from the same entries, with discussions, negotiations and documents.	Rooms
Room Hub	Publish a room for other accounts to subscribe to, or subscribe to one — read-only, always current, connected to your agents’ MCP tools.	My Listings, My Subscriptions
Billing & Metering	What counts as an operation, what each plan includes, and what happens at the limit.	Settings → Usage

Behavior, the Git Repository section, Managed Models and the Pro dashboard pages need a Pro, Teams or Enterprise plan. Security is gated by role rather than plan. Editions has the full comparison.

Getting data in

Migrate imports what your agents already know from mem0, Zep or a Hugging Face dataset. Nothing is written until you have reviewed the mapping. Open Migrate from the dashboard.
Connectors and Integrations bring external events — GitHub, Slack, Linear and others — into the record. See Connectors and Integrations.
Instrumentation gets every run from an agent you operate into the record with steps, LLM calls and cost. See Instrument your agent.

Product

Getting Started

Guides

Self-Hosting

Core Concepts

Pro Features

Making the loop recursive

Seeing what the loop knows

Running it as a team

Getting data in

​Making the loop recursive

​Seeing what the loop knows

​Running it as a team

​Getting data in

Making the loop recursive

Seeing what the loop knows

Running it as a team

Getting data in