Agent Library · By function · Contracts · Obligations · Audit
Every obligation you signed, still tracked.
Mid-market companies sign contracts faster than they track them. The renewal that auto-extended, the SLA nobody monitored, the indemnity nobody re-read. These agents read and remember. They do not advise, and they do not approve.
Contracts are read once
At signing. Then filed. The obligations inside them are managed from memory.
Renewal dates surprise you
Auto-renewal clauses trigger because the notice window passed unnoticed.
Review depends on availability
Whether a contract gets a careful read depends on how busy the reviewer was that week.
The agents
5 agents. Each one earns its autonomy.
Agent
What it does
Autonomy
Human gate
Contract Analyser
Extracts terms, obligations, liabilities and unusual clauses, flagged against your own standard positions.
Review
Legal judgement always human
Obligation Tracker
Turns signed commitments into tracked, dated obligations with owners and reminders.
Auto
—
Renewal Watch
Watches every notice window and renewal date, escalating before the decision window closes.
Flag only
—
Policy Conformance
Checks proposed documents and actions against your internal policies and flags the divergence.
Flag only
—
Audit Trail Assembler
Assembles the evidence pack for an audit or a client security review from your existing records.
Review
—
Autonomy key — Auto: executes above 95% confidence. Notify: executes at 85–95% with notification and undo. Review: always queued for a human. Flag only: never acts, alerts a person.
These agents are not a substitute for counsel. Every output is an input to a human legal judgement, and the system is built so it cannot be anything else.
What changes
Three shifts, not a feature list.
Obligations get managed, not remembered
What you promised becomes a tracked list rather than an act of collective memory.
Renewal becomes a decision again
You choose whether to continue, instead of discovering that you already did.
Review quality stops varying by week
Every contract gets the same first-pass read regardless of who was busy.
The eval bench
What these agents have to pass before they touch real work.
- Clause extraction accuracy against 50 contracts already reviewed by your counsel
- Recall on unusual-term detection — measured on a set with known planted deviations
- Zero legal conclusions in any output; the agent flags and cites, it does not advise
Written against your data, not ours. The bench is built during Blueprint and runs on every deploy from then on.
Related
Where this connects.
See it on your Google Drive.
A Blueprint maps which of these agents pays off first in your exact setup — and which to leave alone.
Apply for a Blueprint