One engine behind every machine that reads your code.
vexp is the code-intelligence layer your automation calls. You build the agent, the bot, the pipeline; vexp gives it the context and the cross-repo impact.
Impact-aware PR review
Your review agent comments on what a change actually breaks across services, not just the diff. vexp resolves the blast-radius and pulls the callers, so the model reasons with real context.
Log to code triage
Turn a stacktrace or error signature from your logs into the owning code, and the team behind it. Cut MTTR by resolving alerts to the exact functions involved.
Context for your agents
Give every internal agent precise, type-accurate codebase context before it edits, across hundreds of repos. The retrieval layer your agents need, that you would otherwise build.
What the engine gives your machines.
The intelligence is the product. Each capability is a call away; you wire it into your own agents and pipelines.
Cross-repo blast-radius
Ask what a change affects and get the answer across your whole fleet: callers, dependents, and the services that break, not just the file in the diff.
Type-accurate resolution
Precise call graphs from real language servers, run server-side inside your index build, with no editor in the loop. Sharp on C#/.NET and 25+ more languages.
Grounded retrieval
From a natural-language task, a stacktrace, or a feedback note to the exact code that matters, bounded to a token budget, ready for your model.
Dual-source indexing
GitHub and Azure DevOps from day one, kept fresh on every push via webhooks, with a nightly safety-net sweep. Built for a mid-migration reality.
Machine-to-machine
A clean REST surface for agents, pipelines, and CI: concurrent, burst-tolerant, with readiness and metrics. You build your own facade on top.
On-prem local model
Intent and compression run on a model inside your network. No data egress, no per-call cloud bill, no third party in the path.
Hundreds of repos. Bounded memory. Your network.
Memory tracks the active working set, not the total repo count, so the footprint stays flat as the fleet grows. Add a worker to scale, not a bigger box.
Measured on a diverse multi-language sample; your monorepos are sized against your own repos during a technical validation.
Typed clients. Build your own facade.
A thin, typed client for Python and TypeScript over a clean REST surface. You wrap it in the REST or MCP facade your agents and dashboards expect; the engine stays yours, self-hosted.
from vexp_sdk import VexpClient
vx = VexpClient("http://gateway.internal:7777", token=TOKEN)
# Blast-radius for a PR, across every repo in the fleet
impact = vx.pr_impact(repo="payments-svc", files=changed_files)
for hit in impact["impacted"]:
print(hit["repo"], hit["fqn"]) # who breaks downstream
# Grounded context for an agent or a log line
ctx = vx.run_pipeline("why does checkout fail", repos=["payments-svc"])
for p in ctx["pivots"]:
print(p["repo"], p["file_path"], p["score"])Start free. Scale to your fleet.
The SDK is a separate, machine-to-machine line, priced by capacity rather than per seat. Begin on the free dev tier; the enterprise capacity license is scoped to your repos, SLA, and deployment.
Dev
Evaluate and small automation
- Up to 5 repositories
- 30 calls / day
- Single worker
- Self-hosted
- Community support
Startup
Scaling automation
- Up to 10 repositories
- Unlimited machine-to-machine calls
- Up to 2 workers
- Webhooks and dual-source
- Email support
Enterprise
Fleet scale, on your infra
- Hundreds to thousands of repos
- Unlimited workers
- Cross-repo blast-radius, fleet-wide
- Service-level agreement (SLA)
Give your machines a map of your code.
Tell us about your fleet: repos, sources, and the agents you are building. We will scope a technical validation on your own infrastructure.
Self-hosted · Your code never leaves your network · Machine-to-machine