For agencies and consultants
You are about to own somebody else’s codebase.
AI made the code cheap to write. The team that wrote it is not the team that has to keep it running — and increasingly that team is you. Forge reads a repository deterministically and produces evidence you can put in front of a client.
Client’s repo is private? npx @ambera/review-tax read — runs on your machine, sends nothing anywhere.
Where it fits
Four moments in an engagement.
Nothing here is a new product surface — it is the same reading, read by a different audience at each step.
Before you quote
Read the repository before you price the work.
Point the free read at any public repo, or run the CLI against the client’s private one on your own machine — no account for them to create and nothing uploaded. You go into the scoping call with counts instead of an impression.
In the pitch
Hand them a reading, not a slide.
One reading composes into a document with its own coverage section — what was read, and what that leaves unanswered — shared by an unguessable link that you can revoke. It says what it measured, which is why a sceptical CTO can check it.
During the engagement
File the fixes where their team already works.
Export findings into their Linear, Jira, GitHub Issues or ClickUp. Each issue carries what was measured, the example paths, and a machine-checkable criterion for what would clear it — so the work is reviewable by someone who was not in the room.
At handover
Show what moved, measured the same way both times.
The closing report compares the window's first reading against its last, and grades what you filed — including where a finding did not move. That is a harder document to write and a much better one to be handed.
Before you buy
What this will not do for you.
An instrument you put in front of a client is only worth having if it can say something you did not want to hear. These are the places it will.
It grades your work, not just their code
A filed action that did not move reads as still observed, and a reading that could not run its detector reads as unreadable rather than resolved. If you would rather a report only ever said good things, this is the wrong instrument.
It never says the fix caused the change
Two readings establish that something moved, never why. Forge reports the movement and leaves the causal claim to you — which is the version a client can trust when you do make one.
There is no agency plan
Each engagement is a Launch Audit: one repository, 90 days, $69 at launch pricing and $99 after. Buy one per client. No volume tier, no white-labelling, no reseller programme — if any of those appear, they will appear here first.
The read is tiered by language, and says which
Deepest on React, TypeScript and JavaScript; language checks for ten more; tree-level for the rest. Every reading names the languages it has no detectors for, so a short list on a Scala service is never mistaken for a clean one — including by your client.
The whole method is published, including where the read is wrong — forge.ambera.app/methodology. Read it before you put a number from this in a proposal.