Agents do the work.
People stay in control.
The record shows every step.
When code an agent wrote breaks production, see which change and which agent session caused it. A person approves the fix, RepoOps measures whether the error stopped, and the lesson reaches the next agent before it edits that code again.
Works with Claude Code, Cursor and other agents that read CLAUDE.md or AGENTS.md. Approval through PromptCompute is switching on, and one sign-in across the products, PromptReports ID, is switching on.
- 1Error traced to a changeDeploy, commit and route, each link marked measured or inferredMeasured
- 2Agent session foundLinked by the Session-Id trailer on the commitExact
- 3Rollback approved by a personNamed approver, signed decisionApproved
- 4Recovery watchedEach success criterion met, not met or not measuredMet
- 5Lesson for the next agentA person reviews it before agents read itIn review
From a production error to a tested lesson.
When code an agent wrote breaks in production, RepoOps follows it back to the change and the session, a person approves the fix, and the result is measured before anything is learned from it.
Agent session
The session that wrote it, linked by a commit trailer or by time and files.
On repoops.ai →One tool call. A person decides.
When an agent wants to deploy, roll back or write to memory, it calls PromptCompute and waits. The right person approves in Slack, email or the web, and the agent picks up where it left off, even days later. Each decision is signed.
request_approval({
action: "code.deploy.production",
args: { service: "checkout", version: "3f2a91c" },
reason: "Roll back the change that broke checkout"
})3f2a91c- Asked by
- Claude Code, session linked
- Change risk
- High broke checkout twice before
- Cause
- Named by blame at the deployed version
- Expires
- in 24 hours, then rejected
Start with one. Connect the rest when you need them.
Each product works on its own and has its own site. Connected, they share one sign-in, one approval inbox, one memory and one public record.
Four things you get once, for every product.
One sign-in
PromptReports ID carries your organization and role into each product.
Built in five, switching onExample: An owner who signs in to RepoOps links the team and connects approvals in the same step.
One approval inbox
Deploys, rollbacks, memory writes and customer messages wait for the same people.
Built, switching onExample: RepoOps sends the change risk with each deploy it asks PromptCompute to run.
One memory
Lessons, decisions and past errors that every agent reads before it acts.
BuiltExample: A bad outcome in PromptCompute starts a lesson in RepoOps.
One public record
Signed provenance and corrections anyone can check later.
Built, off by defaultExample: Each post on this site has a signed record on AI Source of Truth.
What the products do together.
Each workflow runs across several products, and an agent can run it with a person approving where it matters. Every plan includes every workflow. Each shows its true state.
Launch with proof Partly built
Docs verified, what shipped traced, the announcement approved, and each claim recorded with a badge.
ContextFile, RepoOps, PromptCompute, AI Source of Truth
Customer escalation to recovery Partly built
A promise at risk opens a case. The fix and the customer notice are approved together, recovery is measured and the promise kept.
CommitOps, RepoOps, PromptCompute, Sales as Code
Cost per outcome Partly built
Agent spend per recovered incident, kept promise and won deal, with budget approvals.
RepoOps, PromptCompute, CommitOps, Sales as Code
A trust packet a buyer can check Partly built
One signed export of approvals, incidents and recoveries, promises kept and corrections, each item checkable.
PromptCompute, RepoOps, CommitOps, AI Source of Truth
What an agent can do here.
promptreports.ai is public, so an agent needs no token to read it. /family.json gives each product's state, each workflow's state and the evidence for each line, and /llms.txt gives the products, workflows, pricing and posts as plain text.
From a forecast to a promise kept.
The same contract runs for the go-to-market team. A forecast or a proposal is backed by evidence, a person approves what reaches a customer, CommitOps tracks the promise, and the code that keeps it is traced by RepoOps like any other change.
Forecast
Outreach drafted and a forecast kept, with the evidence behind each.
Sales as CodeOn salesascode.com →Approval
An outbound send, such as a proposal, goes to a person before it reaches a customer.
PromptComputeOn promptcompute.ai →Delivery
The pull requests that keep a promise show it, with their change risk and agent cost.
RepoOpsRepoOps page comingOutcome
Forecasts graded every week against what happened. A broken promise opens a case.
Sales as Code, CommitOpsOn commitops.ai →Lesson
A release that forecasts worse is built to file a lesson in the shared brain, reviewed like any other.
RepoOpsOn repoops.ai →Per user, by workflow requests.
Every workflow in the library is included in every plan.
100 requests
- 100 workflow requests a month
- Every workflow in the library
500 requests
- 500 workflow requests a month
- Every workflow in the library
Enterprise
- More than 500 workflow requests a month
- Every workflow in the library
Check our work, not our logos.
Every post on this site has a signed record on AI Source of Truth. Anyone can check that the post they read is the one we recorded.
PromptReports LLC, an Illinois company. A person reads and answers every message.
Build notes.
Launch with proof: each launch claim tied to the change that shipped it
How the Launch with proof workflow checks the docs against a release, puts the announcement in front of a person, and records each claim so anyone can check it later.
Read →Customer escalation to recovery: one case from the complaint to the kept promise
How the escalation workflow will join the customer's report, the error, the fix and the deal into one case, with the fix and the customer notice approved together.
Read →Cost per outcome: what a recovered incident, a kept promise and a won deal cost in agent spend
How the Cost per outcome workflow will join agent spend to the results it produced across products, and hold spend over budget for a person.
Read →Start with the code loop.
Pick a plan and we set up your organization. Questions first? A person answers every message.