What IonWarp is

IonWarp is a repository-quality agent. Connect a GitHub repository and every pull request gets a review: a lineup of reviewers reads the diff, each finding is posted inline on the PR and kept in a ledger until it is fixed or dismissed, and the repository as a whole is scored on how ready it is for the people and coding agents working in it.

Everything IonWarp does works the same from the product and from the API with one key: the same objects under the same rules. An agent running a task cannot change its own project's run mode (Control Plane).

How it fits together

• A project is a repository. Creating one names the repo and its GitHub App installation; setup: true runs repository setup in the same call.
• A review is a task. POST /reviews with {"pr": 42} starts one and answers with the id you poll; reviewers are strings — "code-review", "security-review@google/gemini-3.8-flash" — and omitting them lets the planner pick.
• Findings are a ledger. Each one is an object with a severity, a path and a line, kept open across re-reviews until a push outdates it or a reviewer replies wontfix:.
• Scorecards are the repo-wide rubric. Checks, levels and quick actions that turn a finding into a proposed fix.

Get started

• Connect GitHub — install the app; reviews start on their own. This is checklist step 1.
• Build with AI — one call creates the project and runs setup.
• API quickstart — from a key to a reviewed pull request in six calls.

Read what reviews found

List findings — GET /api/v1/findings

curl "https://ionwarp.com/api/v1/findings?related_pr=141" \
  -H "Authorization: Bearer $DASH_API_KEY" \
  -H "Project-ID: pr_acme"

Show me the open findings on PR 141

data_query {
  "collection": "findings",
  "where": {
    "related_pr": 141
  }
}

{
  "object": "list",
  "data": [
    {
      "object": "finding",
      "id": "obj_123",
      "related_pr": 141,
      "title": "Unbounded loop over untrusted input",
      "severity": "P1",
      "domain": "security",
      "reviewer": "code-review",
      "status": "open",
      "confidence": 4,
      "consequence": "A client that posts a very large array keeps the worker busy until the request times out.",
      "created_at": "2026-07-25T12:00:00.000Z",
      "updated_at": "2026-07-25T12:00:00.000Z"
    }
  ],
  "has_more": false,
  "next_cursor": null
}

Data retention

Review data is deleted 90 days after it was last updated. Every record a review produces (the task, its findings, scorecards, learnings, screenshots, notes about the repository, support tickets) is marked with its delete date when it is written, and a nightly job deletes whatever is past it. Deleting a project deletes its review data with it. The security page says the same.

Pick your path

Reviews · Reviewers & Models · Repositories & PRs · Scorecards · Skills · API reference