The kind of tool you only notice when it’s gone.
RepoQL runs beside the agent you already use and indexes everything it should know about. Git repositories, folders, storage buckets, live telemetry: it all lands in one graph your agent can search, query, and slice.
keywords turns your agent's words into the names this repo actually uses. explore ranks everything those names touch. Parallel explain calls follow the threads, and composite reads pull the exact slices that settle the question.
Concepts and vocabulary are stored beside the code, so the next session doesn't have to re-learn what this one figured out.
Everything indexed has an address: a URI, short for Uniform Resource Identifier. Your agent uses those addresses to find, filter, and slice precisely what it needs, spending tokens only on what matters.
See the sensesNo inbound port and no VPN. The host dials out over a secure WebSocket, and your instance appears as a remote MCP (Model Context Protocol) endpoint at a stable URL: OAuth in front, your whole team and CI behind it.
In the pipeline, the same host spins up inside the job and indexes your branch: your concepts, coverage, and static-analysis findings (SARIF) in one graph.
Plug the review agent into uplink to find downstream breakages and problems that require the big picture to see.
| Where it runs | On your machine, in a container, or on your own infrastructure, and exactly what does and does not need a network. |
| Addressing | Everything indexed has a URI. Schemes, globs, fragments, and modifiers: the one grammar every tool inherits. |
| The senses | What an agent perceives: every file at three depths, structure without reading, one address grammar over all of it. |
explore | Search the whole index by meaning, ranked by how well each result answers your question. |
keywords | Your rough words mapped to the names this codebase actually uses, with evidence. |
read | One slice by address: a symbol, a line range, a glob. A modifier decides what comes back. |
query | DuckDB SQL over the whole graph: files, symbols, git history, parsed data files. |
explain | A synthesized answer with citations, read from more source than an agent would spend context on. |
import | Another repository, an S3 bucket, a folder, or a SARIF report into the same graph. |
execute | Sandboxed JavaScript for the shape SQL cannot express. |
watch | Run a process under an OpenTelemetry (OTEL) collector; its traces, logs, and metrics become queryable tables. |
capture_concept | Write an invariant into the repository's own memory, so the next session inherits it. |
command | Manage the host: status, configuration, imports, accounts, diagnostics. |
| An unfamiliar codebase | Land in a repository you have never seen and reach the code that matters. |
| Profiling with watch | Query an application's telemetry beside the code that produced it. |
| Keeping what you learn | Turn an expensive lesson into something the next session inherits. |
The same manual ships inside the binary at help:///, where your agent reads it the way it reads your code.
It asks before touching anything, wires the agents you already use, and leaves a single command behind.