The platform

One index. More context for the next decision.

Connect structural and semantic retrieval with retained engineering history. Choose the evidence you need through MCP, CLI or Explore.

From finding code to understanding change.

Structural and semantic search

Search by symbol, text or meaning. Hybrid retrieval combines structural context with embeddings from Azure OpenAI, Gemini, Voyage or Ollama.

Call and contract graphs

Trace indexed callers and detected contract dependencies across repositories. Resolution confidence, framework support and traversal limits matter; a missing edge does not establish that no dependency exists.

Compiled architecture context

Retrieve module briefings, scope packs, symbol context and contract maps assembled during indexing. Deterministic retrieval needs no additional LLM generation call; indexing, embeddings, hosting and subsequent agent reasoning still use resources.

Decision history and rationale

Inspect retained commits, linked work items, reviews and source-authored rationale. Historical queries filter retained evidence; they do not reconstruct historical source code or call graphs.

Signals that support engineering judgment.

Change and review context

Review a bounded pack of affected symbols, callers, work items, reviews and contract crossings. Static test reachability suggests paths to inspect; it is not executed test coverage.

Architecture and change history

Inspect structural communities, module lineage, regression provenance and areas undergoing rewrites. Similarity signals help investigate duplicate implementations across repositories.

Contribution evidence

Optional ownership views describe recent contributions and concentration within retained history. They do not measure expertise or employee performance.

The repository-local self-audit workflow can produce calibrated technical-debt findings and controlled issue drafts. Human-reviewed refactor drafts remain an exploratory roadmap item.

Bring context into your workflow.

MCP, CLI and Explore

Give compatible AI clients access over MCP, query from the first-party CLI, or browse through Explore. Confirm the supported installation route for your release; public CLI package distribution is on the roadmap.

Pull-request evidence

GitHub annotations can add impact and historical context to pull requests when enabled. Equivalent review workflows for other source providers are planned extensions.

Shareable reports

Export self-contained HTML graph reports and onboarding context. Optional Teams and Slack integrations need their own identity and access review before use with a restricted corpus.

Connect the tools you already use.

Source control: GitHub, GitLab, Bitbucket and Azure DevOps Git / TFVC.
Work items: Azure DevOps, Jira, Linear, GitHub Issues, GitLab, ClickUp, Asana, Monday and Shortcut. Available context depends on configured ingestion and retained links.
AI clients: MCP-capable configurations of Claude Code, Codex, Cursor, OpenCode, Pi, GitHub Copilot, Windsurf, Continue, JetBrains AI and Antigravity CLI. Transport and authentication support vary by client.
Languages: Symbol extraction for 14 languages, with different parser and framework coverage. Read the capability table.

Operate the index

Freshness is part of the answer.

Durable queues, worker heartbeats, retries and recovery paths help operators deal with interrupted jobs. Health, queue and drift views expose signals to investigate.

Incremental indexing reuses unchanged work where supported. Graph persistence and cross-file processing vary by provider; some stages still scan more broadly. Check the indexed revision and job state before relying on a result.

Read the operating considerations

See Engram in context

Bring a real codebase question.

We’ll use a 30-minute scoping call to understand your repositories, tooling and evaluation goals.

Book a scoping call