One ledger. Three products.
Runledge AI
Live, expandingRunledge AI reads the ledger. Review recorded activity and ask questions about your ledger.
Every action the Gate decided is still here
Runledge AI reads stored proposals, decisions and reported outcomes. Signed action receipts attest to action identity and recorded time, not the contents of these records.
1,248 actions on the ledger
The Gate writes it. This is the reading. Runledge AI works from the same records your auditors will.
Ask Runledge AI about your ledger
Ask a question to filter recorded activity and inspect matching actions. Runledge AI translates your question into ledger filters; it does not verify a provider outcome or authorize an action.
You ask
Which payouts went to payees whose bank details changed this month?
3 payouts — answered from the records below, nothing else
These payouts predate the proposed rule — they are the history the proposal was learned from.
release_payout · $6,480 · bank details changed 2 days before · Held, then approved
release_payout · $3,940 · bank details changed 5 days before · Held, then approved
release_payout · $12,075 · bank details changed 1 day before · Allowed
Every answer cites its actions. If the ledger doesn't say it, Runledge AI doesn't claim it.
Review what your agents recorded
Runledge AI reviews the last 30 days of activity for uncovered patterns and unusual volume. Investigate a finding and open the actions behind it.
Runledge AI review · Last 30 days
3 payouts followed a bank-detail change within 7 days. No rule covers this sequence today.
Investigate · View actions
support-agent issued refunds at 3.2× its daily baseline on 4 days this month.
Investigate · View actions
A review is a reading, not a verdict. Findings open the actions behind them — you judge what they mean.
Turn findings into rule proposals
A finding can become a proposed rule, reviewed alongside its evidence. You remain responsible for every policy change.
Pattern spotted — 3 payouts after bank-detail changes
Proposed rule
Hold any payout to a payee whose bank details changed in the last 7 days.
Seen 3 times this month in the ledger. Two were held by hand and approved; one was allowed.
You decide — it never changes a policy on its own.
The ledger, read back as a rule. A pattern you kept catching by hand, written down as a proposal.
Nothing changes without you
A proposal waits in a queue, like a held action waits for its approver. Accepting one creates a rule in shadow mode only — promotion to enforcing is a second, separate decision.
Hold any payout to a payee whose bank details changed in the last 7 days.
Proposed from 3 ledger records · starts in shadow
In shadow. — It will report what it would have held. Enforcing is your second decision.
The AI is gated too. Runledge checks agents before they act — and its own suggestions wait for a person, exactly the same way.
Runledge AI, in detail
Read — ask and inspect3 capabilities
Questions over your own actions, answered as validated ledger filters plus the matching records.
Read the stored checks and deciding rule without asking a model to invent an explanation.
A 30-day reading of your ledger: uncovered patterns and unusual volume, each with its evidence.
Review — find patterns4 capabilities
An agent acting on a system no rule mentions is itself a finding.
Approvals are data too: rubber-stamps and always-rejected rules both say something — and when a rule's holds are always approved, it proposes a recalibration: a scoped threshold change with backtest evidence and a recheck date.
Some risks only exist as a sequence: a bank-detail change, then a payout.
Model spend per agent, trending — the reading behind the Gate's spend breakers.
Decide — proposals and checks3 capabilities
Patterns become proposed rules in a queue — accept creates a shadow rule; promotion is a second human act.
Compare current and draft rules against bounded action history before deciding to apply a change.
Compare delayed would-be outcomes in the Shadow Pilot Report before choosing to enforce a rule.
Manage — review allowance2 capabilities
Team includes 10 reviews a month; Growth 100; Enterprise unlimited. Developer and Pro do not include Runledge AI.
It never changes a policy, never authorizes an action, never verifies a provider outcome.
Included with your plan
Runledge AI is part of the Gate plans — there is no separate AI bill. Asking questions over your ledger is included from Team onward; reviews are the metered unit.
Full pricing →Questions? The FAQ covers reviews, allowances, and what Runledge AI can and cannot do.
What's your scariest agent action?
Test it in shadow mode first.
Start free — Developer and above — 10,000 actions every month. Start in shadow mode, then block actions when you trust the rules (or hold for review on Pro and above). Shadow is free during the 14-day shadow pilot for each workspace. After the 14-day shadow pilot, shadow Held, Blocked, or Adjusted outcomes count toward your plan like enforcing mode does.
Start free