XRevery
Log in Sign up free
Workflows & agents

Pick a pipeline. Or let the Coordinator compose one.

A workflow decides which specialists touch the work and in what order. Small changes should not pay for a full review cohort, and a risky migration should not skip one.

Direct
fastest, cheapest
implement open PR
One-shot changes: a config fix, a version bump, a typo in a public doc.
Light
balanced
spec implement verify open PR
Contained work where you still want a written plan and a verification pass.
Standard
full cohort
analyse spec review test implement verify open PR
Production paths, cross-service change, anything you would want reviewed twice.
Composed
coordinator built
graph for this task + gates you choose
When the task does not fit a template, the Coordinator proposes a graph and shows it to you before anything runs.

The cohort

Narrow roles beat one generalist told to do everything. Each specialist gets the context it needs and none of the context it does not, which is most of what keeps quality and cost in line.

Spec writer

Turns the brief into a change plan with acceptance criteria and a test approach, using repo profiles and the wiki.

Reviewer

Challenges the spec and the diff: missed cases, contract changes, anything that drifts from house style.

Coder

Implements the approved plan in the sandbox, on a branch, keeping to the conventions in the repo profile.

Tester

Writes and runs tests against the repo's own commands, and reports honestly when something does not pass.

Security reviewerOPTIONAL

Turn it on for auth, payments, data handling, or any path you mark as sensitive.

Holistic reviewerOPTIONAL

Looks at the change in the context of the whole system rather than the diff in isolation.

Catalogue

Start from something that works.

Agents, workflows, skills and MCP servers, in verified and community tiers. Install one, fork it, and change it to match how your organisation works.

Browse it after signup
Agents
Specialists with a defined job and a defined context budget
Workflows
Pipelines and gate placements you can reuse across repos
Skills
Procedures for recurring jobs, such as a framework migration
MCP servers
Give agents access to the tools your team already uses

Gates: your checkpoints, not ours

A gate pauses a run and asks a specific question. You get notified, you answer, the run carries on. Place them where your risk actually is, and leave them off everywhere else.

After the spec
Cheapest place to change direction, before any code is written.
On a contract change
Public APIs, error shapes, database schemas, anything downstream depends on.
On sensitive paths
Mark the directories that always need a human before implementation.
At the merge, always
Not configurable. Agents open pull requests; your team merges them.