Repository-wide code search

Greptile indexes a codebase across its files and history so a question can be answered against the whole repository rather than a single file a person happens to have open. A Beam agent sends a developer's question, such as where a particular function is used or how a module fits together, to that index and returns the passages Greptile finds most relevant. This saves a developer from searching through the repository by hand for context that already exists somewhere in the code. Where the question touches a design decision or a change that has not been made yet, the agent leaves that judgment to the developer rather than proposing one itself.

API access for custom tooling

Beyond its own chat interface, Greptile exposes an API so a team can build the repository search into their own internal tools rather than only using Greptile directly. Where an account has connected that API, a Beam agent can query it as part of a larger workflow, such as pulling relevant code context into a support ticket about a bug or into a code review comment. This keeps the context tied to whatever tool the team already works in instead of requiring a separate lookup. Any query that needs an opinion about how the code should change, rather than a description of how it currently works, is left for a developer to answer.

Change-aware indexing

As a repository changes, Greptile's index needs to reflect the current state of the code rather than an older commit, since a question about a function that was recently rewritten should return today's version, not last month's. Where the connector and account permissions allow it, a Beam agent checks whether the index has been refreshed since a repository's last update and can flag a stale index rather than silently answering from outdated code. This matters most on repositories with frequent commits. If the agent cannot confirm the index is current, it notes that uncertainty when it returns an answer, so a developer knows to double check before relying on it.