LogThread indexes your repository into a live local graph of files, symbols, APIs, queries, schema relationships, architectural boundaries, and runtime evidence. That context stays available through the local UI, CLI, and MCP so humans and coding agents can work from the same grounded view of the system without hosted login, seats, or subscription checks.
Free to use as an open-source local project. Code, graph data, prompts, logs, and schema stay on your machine unless you explicitly configure an external provider.
LogThread works best when teams can move between code understanding, change planning, runtime investigation, and assistant setup without rebuilding context from scratch.
Browse files, symbols, APIs, queries, communities, process groups, and request paths in the Explorer with graph briefs and focused subgraph queries.
Use Change Center and Context Studio for diff impact, implementation briefs, edit contracts, trust signals, and compact assistant handoff packets.
Query supported log providers, cluster repeated patterns, reconstruct journeys, and correlate runtime evidence back to likely code paths.
Add PostgreSQL, MySQL, SQL Server, or SQLite connectors so LogThread can align code-level SQL understanding with real schema state.
Install repo-aware MCP config into supported local assistants, or copy the generated config for manual setup when auto-detection is not available.
Run the lightweight host in the background, open the local workspace in the browser, and share the same host with HTTP MCP clients.
These are current product capabilities, not future roadmap promises.
Graph exploration, call-path tracing, dependency analysis, dead-code review, coupling analysis, workspace mapping, and architecture-focused summaries.
Working-diff analysis, risk review, implementation briefs, edit contracts, trust reporting, and context-cost guidance for leaner assistant handoffs.
CLI and browser UI, repo-aware MCP access, local host management, assistant installs, and optional runtime or database connectors that stay tied to the active workspace.
Interactive graph view, graph brief, shortest-path style tracing, focused graph queries, and exportable HTML reports.
Impact analysis, agent-safe contracts, workspace boundaries, trust review, and implementation planning around the current repository state.
Runtime query workspace, provider connectors, pattern clustering, journey reconstruction, and code correlation for supported log sources.
Repo-aware MCP setup, compact context packaging, context-cost reporting, and a shared local host for multiple assistant clients.
Question: Where does this route fan out after auth?
✦ Graph brief: entry surface touches route → service → repository
✦ Surprising link: shared schema dependency crosses two communities
✦ Suggested follow-up: inspect downstream write path and tests
Next: send a compact packet to an assistant instead of full files
$ logthread change-risk
Working diff: 7 files · Impact path detected · Trust notes ready
Assistant brief: compact packet preferred over raw file dumps
Runtime clue: latest error slice maps back to ingestion worker
Suggested follow-up: open Context Studio or Runtime view
Start with the local CLI and browser UI, index the repository, then add runtime and database connectors only where they help.
Run pip install logthread
Run logthread generate --full in the repository you want to understand.
Add runtime and database access from the UI when you need schema sync, SQL checks, or runtime correlation.
Use logthread ui, CLI workflows, or assistant installs through MCP.
LogThread is designed to sit on top of your current stack, not replace it. Point it at your existing observability systems and databases when you want more grounded answers.
Connect supported log providers such as Grafana Loki, Datadog Logs, Splunk, Sumo Logic, CloudWatch, Elasticsearch, or New Relic.
logthread logs connect --provider loki --name prod
Add PostgreSQL, MySQL, SQL Server, or SQLite connectors so code-level SQL paths can be checked against real schema state.
Useful when schema drift or SQL validation matters as much as code navigation.
Use request IDs, traces, services, routes, and schema hints so LogThread can connect failures back to likely code, dependency, and data boundaries.
The indexing, UI, change analysis, and assistant workflows remain local to your machine.
Start with the local CLI and browser UI. Teams can standardize shared setup docs and assistant configs without forcing contributors through hosted accounts, seat assignment, or subscription checks.
Install the package, index a repository, then open the local workspace:
logthread generate .
and logthread ui.
Best for quick local understanding, code review prep, and assistant grounding.
Share recommended setup commands, MCP client config, runtime connector conventions, and project names in your repo docs. Each developer keeps their local graph and credentials under their own machine.
Use this when multiple developers or AI assistants need the same repo-aware language without central billing.
The product is strongest when teams move from understanding to action without losing repo context.
Start from a compact graph brief, follow focused subgraph queries, and answer architecture questions without dumping raw files into prompts.
Build implementation briefs, edit contracts, and compact context packs that make assistant handoffs smaller and more repo-aware.
Check SQL against indexed and connected schema information, then follow the path back into the code that builds or executes it.
Review diffs, inspect blast radius, surface trust gaps, and run security-aware analysis before pushing an assistant-generated change further.
Understand nested repos, workspace boundaries, live diffs, and the likely effect of local edits before they spread across the codebase.
Keep the browser UI and shared MCP endpoint available through the lightweight local host instead of restarting tooling every session.
Use LogThread as a developer tool in your own repositories without a hosted account, paid seat, or subscription gate.