Context Goblin reviews pull requests. Register a second repo and it maps how your services are wired together, so the reviewer sees everything downstream that has an opinion about the change. You pick the reviewers with customizable subagents, prompts and MCPs. Install the GitHub App and your next PR gets reviewed. Free and instant for the first signed-up teams.
This is the 2nd launch from Context Goblin. View more
Context Goblin
Launched this week
The quality check for AI-built software
An AI code reviewer with a senior engineer's memory of your codebase. It knows who calls what and which service depends on it, so it reviews the change, not just the diff.
Context GoblinCode review that knows what your PR breaks in other repos
Launched on August 26th, 2026
5.0
Based on 1 review
Review Context Goblin?
Reviews
Most Informative
Maker
📌
Context Goblin started as an internal tool. Coding agents produce far more code
than anyone can review by hand, and every reviewer we tried read the diff without
reading what depends on it.
What's in it today:
- Call graph per repo, so it reviews the change against its real callers
- Cross-repo map: routes, queues, shared tables and packages between your services
- Customizable review agents for your specific needs
- Bring your own MCP servers for ticket and org-specific context
- GitHub App
Happy to answer anything about how the context pipeline works.
Report
The call graph is the right unit. The worst bug we shipped this year was in a file I'd never opened and I approved that PR twice, and no diff-only reviewer was ever going to catch it. What I'd want on this page is the false positive rate, because a bot that leaves a comment on every PR gets muted by week two and then it's worse than nothing. Also dynamic dispatch and reflection are where call graphs go quiet, so does it tell me it couldn't resolve a caller, or does it just say nothing?
Report
Maker
@asadmalik901 Thanks for the interest Asad! I'm fully on your side concerning the noise a bot generates. We very much try to tailor the responses to the verbosity our team needs and for that we use a dedicated agents which decides what to raise and what not. Until now we don't have a detailed number of false positives as it is somehow hard to calculate if a lot of the responses from the Goblin are read by agents as well. Regarding the call graph, we do have an amount of observability and the agent explicitly mentions if an issue is dependent on some other codebase it couldnt resolve.
the cross-repo call graph is the interesting part here, most PR reviewers stop at the diff. how fresh does that map stay as services evolve - is it rebuilt on every push, or something you have to periodically re-sync? asking because a stale map that confidently tells you "nothing downstream depends on this" seems worse than no map at all.
Report
Maker
@galdayan hi Gal, Context Goblin re-syncs periodically, although we are still working on a good solution for big dependency graphs. Currently, you can trigger a sync yourself and with every onboarding of new repos.
The call graph is the right unit. The worst bug we shipped this year was in a file I'd never opened and I approved that PR twice, and no diff-only reviewer was ever going to catch it. What I'd want on this page is the false positive rate, because a bot that leaves a comment on every PR gets muted by week two and then it's worse than nothing. Also dynamic dispatch and reflection are where call graphs go quiet, so does it tell me it couldn't resolve a caller, or does it just say nothing?
@asadmalik901 Thanks for the interest Asad! I'm fully on your side concerning the noise a bot generates. We very much try to tailor the responses to the verbosity our team needs and for that we use a dedicated agents which decides what to raise and what not. Until now we don't have a detailed number of false positives as it is somehow hard to calculate if a lot of the responses from the Goblin are read by agents as well. Regarding the call graph, we do have an amount of observability and the agent explicitly mentions if an issue is dependent on some other codebase it couldnt resolve.
Dial
the cross-repo call graph is the interesting part here, most PR reviewers stop at the diff. how fresh does that map stay as services evolve - is it rebuilt on every push, or something you have to periodically re-sync? asking because a stale map that confidently tells you "nothing downstream depends on this" seems worse than no map at all.
@galdayan hi Gal, Context Goblin re-syncs periodically, although we are still working on a good solution for big dependency graphs. Currently, you can trigger a sync yourself and with every onboarding of new repos.