What's real, what's next.
A ledger of what's live in M9R today, what's being built, and what comes after, in the order it will actually happen.
Live today, linked to the real thing
On the bench, actively being built
Committed direction, honest sequence
What already holds up in court.
Every line below links to the real thing. No mockups, no waitlists.
Evidence levels on every finding
Inferred, correlated, or command-tied. If a run can't support a claim, the record says so.
- Open Memory →
Workspace memory
Confirm what an agent flagged; keep, rewrite, or archive it on evidence. Export what the team remembers to AGENTS.md, CLAUDE.md, or a Cursor rule.
- Open the workspace →
Agent Workspace
Per-agent workspaces, evidence, and approvals for Claude Code, Codex, and Grok Build, in one command center for the custody loop.
- Connect your agent →
The M9R CLI
npx m9r-cli init connects an agent; submit-session seals a run; rules binds the next one. No SDK, no re-instrumentation.
Run A vs Run B
Compare a later run against the sealed baseline and see whether retries and thrash actually fell, described conservatively.
- See pricing →
Billing for paid tiers
Free stays free: up to 2 workspaces, up to 2 connected agents, the full governance loop. Pro is a real Stripe subscription, managed from your own billing portal.
- Connect GitHub →
Real GitHub integration
Push/PR/review activity from GitHub posts live into the channel it's bound to. An agent can propose a pull request; a human decision is the only thing that opens it for real. This is scoped per workspace, so it only ever touches repos you connected.
Channel workflow automation
Message- or schedule-triggered workflows that post, request approval, or run steps in order, behind the same human-decision gate every other path in the product uses.
On the bench right now.
Actively being built. Close enough to name, still too early to link.
Rule health signals
Usage and quality reporting for bound rules, shown only when runs carry real metadata and objective signals.
First-run onboarding
Every empty screen teaches the same one command, so the path from zero to a sealed record takes minutes without a doc-dive.
Next in the chain.
Committed direction, honest about sequence, with no dates we'd have to walk back.
Team workspaces
Shared workspaces with SSO and an audit log, so evidence and rules are governed the same way the code is.
CI gate & PR checks
Block a merge when the run that produced it violated the rules it was bound to. The chain of custody meets the pipeline.
More trace providers
Broader ingestion across agent CLIs and observability formats, so any run can become a record.
Hosted API & SDK
Programmatic sealing and rules fetch for custom agents: the same custody loop, callable from your own harness.
OathExchange
A directory for comparing agents by record-backed run history and workspace fit.
OathLedger
A repo timeline for sealed sessions, approved evidence, and bound workspace rules.
OathGuard
A policy-routing concept for carrying approved repo rules into agent starts.
The shipped column only grows.
Try what's live today. Connecting an agent takes minutes and doesn't need a signup.