Bito AI Code Review Agent

You will like Bito if you are the team on Bitbucket or self-managed GitLab that every other reviewer treats as an afterthought; it covers GitHub, GitLab and Bitbucket including the self-managed variants, which for a regulated shop is the whole decision. The comments are dense, and you will spend week one turning categories off until the signal shows through. After that it is a steady second reader.

Pick it for a mixed-host shop or an on-premises Git server. Pick CodeRabbit if you are on GitHub and want the larger ecosystem of integrations.

The risk is the meter. Each seat includes 5K reviewed lines a month and every extra thousand costs $5. Lines are the wrong unit: a formatter run, a generated client, or a lockfile bump consumes quota and produces nothing worth reading, so the bill tracks churn rather than value, and the months with the most refactoring are the months with the biggest overage.

Exclude generated paths before the first invoice, because the tool bills what it sees. What it does right: fbinfer and Dependency-Check findings are folded into the same review, so part of every comment thread is deterministic and reproducible.

The documented context model has three layers: 1. symbol indexing, 2. abstract syntax trees, 3. embeddings, feeding specialized reviewers for performance, structure, security, optimization and scalability. That is more than diff-only tools do, and the layering is principled: symbols for precision, embeddings for recall, trees for structure. The same page claims merges sped up 89%, regressions cut 34% and 87% human-grade feedback, with no cohort, baseline or rater described.

Claimed, not demonstrated. What would change the assessment is a description of who was measured against what. The observation: two-digit precision on a number with no denominator.

Bito shipped in January 2023 as an assistant and pivoted into review when that market opened, which reads as adaptive rather than accidental, and adaptive is what a small company needs to be. The tell is AI Architect: a codebase knowledge graph sold usage-based to coding agents, a second product launched before the first has pricing power, which splits a small roadmap two ways.

Likely acquirer: Atlassian, given the Bitbucket coverage rivals ignore and a buyer that needs a reviewer of its own. Position: small; watch which product gets the roadmap and whether the graph finds its own customers.

The demo is a review comment that cites the right file. Procurement: Team is $12 a seat on annual, so $720 a month for sixty, or $1,200 on Professional, the tier with Jira and Confluence. Self-hosted is a $5 per seat add-on, another $300, which is cheap for a deployment security will prefer. SSO and SAML sit in Enterprise, so the cheap tiers assume shared logins. SOC 2 Type II is stated.

Onboarding is a Git app install and a config file per repo. Approved with conditions: an Enterprise quote for SSO, and self-hosted if the questionnaire asks where the diff goes.

No source, no BYOK, Claude chosen for me, and nothing runs on my box. The handles are real, though: the CLI lives at gitbito/CLI on GitHub, the review agent is customized through a bito-cra.properties file I can commit next to the code, and AI Architect exposes the codebase knowledge graph as an MCP server, so Claude Code or Cline can ask it where a symbol lives without me pasting paths.

That last part I would actually use, since a graph I can query from my own agent is worth more than a reviewer I cannot read. The rest is rented. Grudging respect for the properties file.