The Context Layer for AI Engineering
Less than 20% of companies are getting real ROI from their AI coding tools. OpenTrace fixes that — giving your AI the system context it needs to work safely and confidently across your entire codebase.
Index public and private repos, analyze local codebases, build rich relationship graphs, and explore architecture — giving AI a complete picture of how your code is structured.
Chat directly with your codebases, explore services, understand system relationships, and surface hidden dependencies — so your AI tools answer questions instead of making assumptions.
Analyze pull requests, review proposed changes, understand affected components, and map potential blast radius — before AI-generated code ever reaches production.
Deep connectivity via Claude, OpenCode, and MCP (Model Context Protocol) — powering project-based workspaces with true engineering context across every AI tool in your stack.
Your AI coding tools are only as good as what they know about your systems. OpenTrace connects them to a live understanding of your codebase, dependencies, ownership, and architecture — so they stop guessing and start delivering.
OpenTrace continuously maps your repositories, services, and dependencies in real time — so your AI always has an accurate picture of how your systems actually work, not how they worked six months ago.
Before any AI-generated change goes to production, OpenTrace maps the full blast radius — every downstream dependency, every affected service. Your engineers can review with confidence, not anxiety.
Engineers and AI tools can query your entire codebase in natural language — ownership, dependencies, data flows, architecture decisions. Answers in seconds, not hours of digging.
Automatically surfaces architectural anti-patterns, circular dependencies, and risky changes — before they reach production.
A real-time map of your running infrastructure — so AI always operates on what's actually deployed, not an outdated diagram.
Code, runtime, infrastructure, ownership, and history — unified in one shared context that every AI tool in your stack can draw from.
claude code
chargeCard(), what breaks?
opentrace.find_usages…
chargeCard() is called from 12 sites across billing-api, order-service, payments-svc, and notification-svc. Hot path is POST /checkout (~14k req/s, p99 42ms). Three callers live outside this repo. I'll draft a rename with a deprecation shim.
Install the OpenTrace plugin for Claude Code and give Anthropic's coding agent live access to your architecture graph: dependencies, runtime data, ownership, and history. Stop pasting stack traces. Start asking.
claude plugin marketplace add
https://github.com/opentrace/opentrace
claude plugin install opentrace-oss@opentrace-oss
Three interconnected layers that give AI and engineers a complete picture of your system.
Continuously maps repositories, services, classes, and dependencies in real-time via Git commits. Understands not just what your code is, but how it connects.
// Auto-detected dependency graph
payment-api
├── imports auth-middleware
├── calls user-service/v2
├── queries billing-db
└── emits payment.completedUnifies deployments, traces, logs, infrastructure state, performance metrics, and tickets into a single operational view. See what's actually happening, not what should be.
Links code and runtime data into a unified knowledge graph, providing actionable system insight to AI tools and engineers. The connective tissue that makes everything queryable.
Most AI code reviewers only see the diff. OpenTrace sees the full picture: every dependency, every downstream consumer, every production metric behind the code you're changing.
- const timeout = 5000;
+ const timeout = 15000;
This timeout change affects billing-api, order-service, and notification-svc. Current p99 latency for this path is 3.2s. Increasing to 15s may mask upstream failures. Last incident on this path: 12 days ago (PAY-1847).
OpenTrace plugs directly into Claude Code, Cursor, Copilot, and more via MCP — enriching every AI tool in your stack with deep system context, instantly.
Indexes code, PRs, and dependency graphs across your repositories in real time.
Maps deployments, services, and infrastructure topology as they run.
Connects spans, traces, and logs back to the code and services that produce them.
Surfaces messages, tickets, and decisions, giving AI agents direct access via MCP.
Large refactors and migrations become data-driven. Know the blast radius before you push, not after the pager goes off.
Eliminate manual code archaeology. Turn days of investigation into seconds with instant system queries.
Identify weak links and failure chains before they cascade. Your architecture becomes a living safety net.
Join the teams already using OpenTrace to unlock the real ROI from their AI coding tools — with full system context, confidence, and control.
No credit card required. Design Partners get direct access to the founders and shape the product roadmap.
OpenTrace is the context layer for AI-assisted engineering. It builds a living knowledge graph of your entire system — code structure, service dependencies, ownership, runtime behaviour, and operational history — and connects it to your AI coding tools via MCP. This gives tools like Claude Code, Cursor, and Copilot the situational awareness they need to work safely and effectively in your codebase.
According to Gartner, less than 20% of companies are getting real gains and ROI from AI coding tools. The core reason is that AI tools lack system context — they don't know your dependencies, who owns what, what has broken before, or what the blast radius of a change is. OpenTrace provides this context, enabling AI to work confidently and safely in your systems.
OpenTrace connects to AI coding tools via MCP (Model Context Protocol). Once connected, your AI tools can query your knowledge graph in real time — understanding service dependencies, ownership, blast radius, and operational history before making any change.
Yes. OpenTrace OSS is a free, open-source knowledge graph tool available under the Apache 2.0 licence. It lets engineering teams index repositories, analyse codebases, and explore architecture — self-hosted on your own infrastructure. A hosted platform with additional engineering agent capabilities is also available.
Situational awareness in AI engineering means giving AI systems the same contextual understanding that experienced engineers carry — knowing which services are critical, who owns what, what has broken before, what the blast radius of a change is, and what operational constraints exist. Without situational awareness, AI tools produce changes that are technically correct but operationally wrong.