Graphite Agent (formerly Diamond)vsTusk
Generated from the two spec rows. Green marks the better value where a spec has a clear direction. Everything else is just different.
Graphite Agent (formerly Diamond)
Graphite · Code review agent
#119
Panel
5.80 spec wins
- Reliability
- 5.5
- Usefulness
- 5.5
- Cost
- 5.5
- Longevity
- 6.5
“Launched as Diamond in March and retired the name by October, a shorter lifespan than some of the pull requests it reviewed.”
Tusk
Tusk AI · Code review agent
#158
Panel
5.73 spec wins
- Reliability
- 5.3
- Usefulness
- 6.3
- Cost
- 4.8
- Longevity
- 6.2
“It has a bot called CoverBot for backfilling coverage on code you already shipped, which is the most honest job title in software.”
Spec by spec
| Spec | Graphite Agent (formerly Diamond) | Tusk |
|---|---|---|
| Architecture | ||
| Category | Code review agent | Code review agent |
| Runs | cloud | cloud, local |
| Platforms | web | macos, linux, web |
| Context window | not documented | not documented |
| Protocols | ||
| MCP client | No | No |
| MCP server | No | No |
| Capabilities | ||
| Runs terminal commands | No | Yes |
| Multi-file edits | No | Yes |
| Git operations | Yes | Yes |
| Browser control | No | No |
| Sandboxed execution | No | No |
| Multi-agent orchestration | No | No |
| Headless / CI mode | Yes | YesGenerated tests are designed to run locally or in CI as a PR check. |
| Models | ||
| Backbone | Anthropic, OpenAI | undisclosed |
| Bring your own model | No | No |
| Local models | No | No |
| Cost | ||
| Pricing model | seat | seat |
| Starts at | $20/mo | $0/mo |
| Free tier | Yes | Yes |
| Bring your own key | No | No |
| Openness | ||
| Open source | No | No |
| License | proprietary | proprietary |
| GitHub stars | n/a | n/a |
Which one would each critic pick
| Critic | Graphite Agent (formerly Diamond) | Tusk | Pick |
|---|---|---|---|
| El Juez | — | — | not enough reviews |
| El Amigo | 6.5 | 6.0 | Graphite Agent (formerly Diamond) — Pick Graphite Agent if your team already stacks pull requests in Graphite, because the reviewer comes with the seat; pick cubic if you only want a reviewer. |
| El Crítico | 6.3 | 5.3 | Graphite Agent (formerly Diamond) — The agent reviews, fixes and updates pull requests with no sandbox in the product, so a reviewer that writes code has nowhere isolated to write it. |
| El Profesor | 5.5 | 5.8 | Tusk — Business context is named as an input and never defined, so the mechanism converting a captured request into an assertion is the part left entirely undescribed. |
| La Inversora | 7.3 | 6.8 | Graphite Agent (formerly Diamond) — Graphite's asset is the stacking workflow and merge queue; AI review is a retention feature priced into a seat it already sells, which is where the pricing power lives. |
| La Jefa | 6.0 | 6.3 | Tusk — Team pricing is $50 per active developer a month, so sixty engineers is $3,000, and the generated tests run as an ordinary check in the pipeline we already have. |
| El Hacker | 3.0 | 4.0 | Tusk — Proprietary, undisclosed model, no key of mine and no weights on my machine; the one thing I get is a command-line entry point I can drive from a script. |
Picks are derived from each critic's own scores. Humans vote on matchups on the duels page.