The modern forge for productivity
Sourcerer is a forge built natively for jujutsu, where review takes the shape of your changes. Whether you like your changes in a stack or even a tree, Sourcerer will adapt.
If you'd simply like to know more, reach out at hi@sourcerer.sh
What we're focusing on
- 01Increasing review productivityReview smaller changes faster, with better context.
- 02Reducing review load from agentsAgentic code output is growing fast, so review capacity has to grow with it.
- 03Scaling your monorepoGive humans and agents the whole codebase to work with, without running into walls.
- 04Auditable review evidenceAudit prep is a simple export of data, not a quagmire of a process.
- 05EU data sovereigntyWe love our data hosted closer to us — maybe you do, too.
Review productivity
The bottleneck was never writing the code
Review queues stall because the unit of review is wrong. Sourcerer takes jujutsu's model at face value: the change is the unit, stacks are normal, and rewriting history is routine rather than an emergency.
Review the change, not the branch
Each change in a stack gets its own thread, its own approval, its own history. Landing the bottom of the stack doesn't invalidate the reviews above it.
Rewrites keep their context
When an author reworks a change, comments follow it. Reviewers see what actually moved instead of re-reading a diff they already approved.
Nested ownership
When a change hits more areas of the codebase, there's no need to try and coordinate 50 groups of owners — you can find the smaller set that helps you move faster.
Agent review load
Machines write faster than people read
An agent can open twenty changes before lunch. A queue that worked when every change had a human behind it stops working the moment most of them don't. Sourcerer lets you set the rules for what can land unattended, then learns from your review history which changes actually needed a person — so your engineers spend their attention where it changes the outcome.
Custom review rules
The experimental corners of your codebase can land on automated review alone, while the parts you can't afford to get wrong keep a human gate. You write the rules and change them as the risk does.
Learning from your feedback
Every review leaves a signal: whether the comments changed the outcome or confirmed what was already fine. Sourcerer reads that history and shows you where the human pass is earning its time.
Monorepo scaling
Built for one big repository, not a thousand small ones
Splitting a codebase to keep the forge responsive is a workaround for the forge. Sourcerer targets the single large repository as the normal case: one history, one review surface, and performance that is our engineering problem rather than your migration project.
Storage is an implementation detail
The repository sits behind an interface we own and can replace. Scale work happens behind the scenes instead of asking you to reshape your repositories for it.
Tools that help teams coexist
Improved ownership management to share and divide responsibility, as well as siloing sensitive code from eyes that should not know what's in there.
Audit evidence
Approval is evidence, not a green checkmark
Reconstructing who approved what (and against which version) from a comment thread is archaeology. Sourcerer records the review as durable evidence at the moment it happens, so the answer already exists before anyone asks for it.
Approvals bind to a version
An approval names the exact change state it was given against. When that state is rewritten, the record says so instead of quietly carrying the checkmark forward.
The trail outlives the discussion
Comments, resolutions and approvals survive rebases, squashes and rewrites, because they attach to the change rather than to a line in a diff that has since moved.
Exportable, not screenshot-able
The evidence for a release is a query you run, not a folder of screenshots someone collected the week before the audit.
EU data sovereignty
European by default, not by upgrade
Where source code lives is a procurement question long before it is a technical one. Sourcerer is hosted in the EU and operated from the EU.
One jurisdiction, no asterisks
Source, review history, metadata and backups stay in EU jurisdiction. There's no region toggle to get wrong, and no control plane sitting elsewhere behind an EU data plane.
Subprocessors stay legible
A short, published list of what we run on and where it runs. Fewer moving parts is easier to answer for than a long compliance narrative.
If your review queue is the bottleneck, we want to hear about it.
Sourcerer is in active development. Join the waitlist and tell us what your review process actually looks like — it's shaping what ships first.
If you'd simply like to know more, reach out at hi@sourcerer.sh