Tenki vs Graphite: AI Code Review Pricing

Graphite and Tenki both review pull requests with AI, but they solve different problems. Graphite is a full PR workflow platform with stacked PRs, a merge queue, a CLI, and an inbox. Tenki is a CI merge gate that does one thing: reviews code and blocks the merge if something's wrong. The pricing reflects that gap in scope, and so does the architecture.

There's also a newer variable: Cursor acquired Graphite in December 2025. That changes the vendor independence picture in ways that matter to platform teams doing toolchain evaluations.

Pricing: per-seat vs. per-review

Graphite uses per-seat pricing with four tiers. Hobby is free but limits AI reviews. Starter runs $20/user/month (billed annually). The Team tier, which is where you actually get unlimited AI reviews plus merge queue access, costs $40/user/month billed annually. Enterprise is custom pricing.

Tenki charges $1 per review. No seat count, no tier gating. The Starter plan includes $10 in monthly credits (10 reviews), and the Team plan includes $100 in credits. Beyond those credits, it's still $1/review.

The math is straightforward. Take a 10-developer team merging about 50 PRs per month:

• Graphite Team: 10 users x $40/month = $400/month ($4,800/year)
• Tenki: 50 reviews x $1 = $50/month ($600/year)

That's an 87% cost difference on the code review line item alone. The gap widens as team size grows because Graphite's cost scales with headcount, while Tenki's scales with PR volume. A 30-person team doing the same 50 PRs/month pays $1,200/month on Graphite Team and still $50/month on Tenki.

The fair counterpoint: Graphite's $40/seat includes more than just AI reviews. You're also getting stacked PRs, the merge queue, the inbox, automations, and team insights. If you'd pay for those features anyway, the effective review cost drops. But if you only want the review, you're subsidizing a suite you might not use.

Rate limits and tier gating

Graphite's Hobby and Starter tiers both include only "limited" AI reviews. The pricing page doesn't publish exact numbers, but unlimited reviews require the Team plan at $40/user/month. That means a small team on the $20 Starter tier can hit a review cap without knowing the threshold in advance.

Tenki has no rate limits. Every review costs $1, period. You can run 5 reviews a month or 500; the unit price doesn't change and there's no tier jump required to unlock more. If your team has a busy sprint week, the cost goes up proportionally. If it's a quiet month, it goes down.

CI integration: advisory comments vs. merge gates

This is the architectural difference that matters most in practice.

Graphite's AI reviews post comments in the PR thread, similar to how a human reviewer would. They surface findings, suggest fixes, and summarize CI results. But the review itself is advisory. It doesn't set a GitHub status check. If branch protection requires a passing status, Graphite's AI review isn't wired into that gate by default.

Tenki runs inside your CI pipeline as a GitHub Action. It posts a status check that can be set as a required check in branch protection rules. If the review finds critical issues, the check fails and the PR can't merge. This is a real enforcement mechanism, not a suggestion. For teams that want AI review to be a gate rather than a comment, Tenki's approach is structurally different.

The tradeoff: Graphite's PR-thread approach is less disruptive. Developers see comments and decide whether to act on them. Tenki's approach is stricter, which can slow things down when the reviewer flags false positives. But if you want a hard gate, you need the CI integration.

Product scope: suite vs. focused tool

Graphite does a lot. Beyond AI reviews, you get:

• Stacked PRs via CLI, VS Code extension, and MCP
• Merge queue (basic on Team, advanced on Enterprise)
• PR inbox with Slack notifications and team insights
• Graphite Chat and AI-generated PR titles/descriptions
• Automations and CI optimization (Team tier)
• Cursor Cloud Agents embedded directly in the PR workflow (newly added post-acquisition)

Tenki doesn't offer any of that. No stacked PR tooling, no merge queue UI, no PR inbox. It's a code reviewer that runs in CI. That's it.

If your team already uses stacked PRs and wants a unified platform, Graphite's breadth is genuinely useful. But if you're running a standard GitHub branch-and-merge workflow and want an AI reviewer bolted onto your existing CI, Tenki's narrow scope is the point. Less surface area means fewer decisions about which features to adopt and fewer vendor-specific workflows to learn.

The Cursor acquisition and what it means for lock-in

In December 2025, Anysphere (Cursor's parent company) acquired Graphite in a cash-and-stock deal valued above Graphite's previous $290 million valuation. As of June 2026, Graphite's contact and sales pages route through cursor.com/contact-sales, and a banner on graphite.com reads: "Cursor Cloud Agents are now in Graphite."

This matters for a specific reason. If you use Cursor as your code editor, Cursor Cloud Agents to write code, and Graphite to review and merge that code, your entire write-review-merge pipeline runs through a single vendor. That's a concentration of control that some platform teams will flag during toolchain reviews.

The concern isn't that Cursor/Graphite will do something nefarious. It's subtler than that. When the same company builds the tool that writes code and the tool that reviews code, the incentive alignment shifts. A code review tool's job is to find problems. An AI coding tool's job is to produce output that passes review. If both are optimized by the same team, there's a structural pressure to reduce friction between them, which might mean fewer findings, more lenient defaults, or benchmarks tuned to the parent company's generation model.

Is that happening today? There's no evidence of it. But the structural incentive exists, and it's worth being aware of.

Tenki has no affiliation with any code editor or AI coding agent. It's an independent reviewer. The company sells CI runners and code reviews, not code generation. That separation of concerns is architectural, not just organizational.

Review quality: the benchmark numbers

Tenki publishes a public benchmark that tests seven AI reviewers against 122 real production bugs, scored by a 3-LLM judge panel. The recall numbers tell a clear story:

• Tenki: 68.9% recall (84/122 bugs caught), 29.9% precision, F1 of 41.7
• Graphite: 3.3% recall (4/122 bugs caught), 50% precision, F1 of 6.2
• Cursor: 32% recall (39/122 bugs caught), 51.3% precision, F1 of 39.4

Graphite's high precision (50%) means when it does flag something, it's usually right. But catching only 4 out of 122 bugs means 118 bugs pass through undetected. For a tool positioned as an AI code reviewer, that's a problem.

Tenki's recall is the highest in the benchmark at 68.9%. Its precision is lower (29.9%), which means more false positives, but for a merge gate the priority is catching real bugs before they ship. You'd rather investigate a false positive than deploy a real bug.

It's worth flagging that this is Tenki's own benchmark. They publish the methodology and invite scrutiny, but it's still a first-party source. Take it as directional rather than definitive.

What Tenki doesn't include

Honest comparison requires listing what you give up by choosing Tenki over Graphite:

• No stacked PR CLI or VS Code extension. If your team relies on stacking, you need Graphite (or a separate tool like ghstack).
• No merge queue UI. GitHub has its own merge queue feature, but Graphite's adds more control and visibility.
• No PR inbox or Slack-integrated notification routing. You're using GitHub's native notification system.
• No AI-generated PR descriptions or chat. You write your own PR descriptions like before.
• No team insights or analytics dashboard. If you want PR cycle time metrics, you need another tool.

None of these are review features. They're PR workflow features. If you need them, Graphite is genuinely the better fit. But conflating "PR workflow platform" with "AI code reviewer" is how teams end up paying $40/seat for a review capability they could get for $1/review.

Which one to pick

Choose Graphite if you want a complete PR workflow suite. If your team uses stacked PRs, wants a merge queue with advanced controls, and values having a PR inbox with Slack integration, Graphite bundles all of that with AI reviews included. The per-seat cost makes more sense when you're actually using the full platform. The Cursor integration is a bonus if you're already a Cursor shop.

Choose Tenki if you want an AI code reviewer that acts as a CI merge gate, costs 80-90% less than Graphite for most team sizes, and isn't tied to any editor or code generation vendor. If your team runs a standard GitHub workflow without stacked PRs and just needs a high-recall automated reviewer that blocks bad merges, Tenki's code reviewer does exactly that.

Note: Graphite's pricing and feature set may change as the Cursor integration deepens. This comparison is based on graphite.com/pricing as of June 8, 2026.