ctx

ctx is a Python CLI and library that recommends a small, relevant set of skills, agents, and MCP servers for a repository or task. It can use your organization's local/private knowledge or the shipped graph. Harness selection for custom, API, and local models is a separate workflow. Recommendations do not install or load tools without an operator action or approval.

Release status: v1.0.20 is the current GitHub and PyPI release; this source tree declares 1.0.21 for unreleased work.

Install

Requires CPython 3.11 or newer. Linux and macOS are the tested host platforms; other POSIX systems are best-effort. Native Windows and PowerShell are not supported. On a Windows machine, run ctx inside WSL2 as a Linux installation.

pip install claude-ctx

Quickstart

From the repository you want to analyze, install the runtime graph and request recommendations:

ctx-init --graph --model-mode skip
ctx-scan-repo --repo . --recommend

ctx-init --graph uses the bundled runtime artifact in a source checkout or downloads the matching release asset for a package install. The full packed wiki is optional; see the knowledge graph guide.

Every clean graph install seeds nine project-owned, MIT-licensed, no-key fallbacks: ctx-python-testing, ctx-python-state-protocols, ctx-python-input-boundaries, ctx-python-api-compatibility, ctx-javascript-testing, ctx-rust-patterns, ctx-typescript, the ctx-python-reviewer agent, and the local ctx-core MCP server. ctx preserves unrelated skill, agent, MCP, and converted-skill content. Runtime-managed harness pages are refreshed from the installed artifact. Installation fails closed if a reserved ctx-* identity, body, overlay, or parent path is unexpected.

Privacy And Telemetry

These controls are available in release 1.0.20 and the current source tree. Telemetry is enabled by default in local_redacted mode. Events are written to ~/.ctx/telemetry/events.jsonl, metrics are written to ~/.ctx/telemetry/metrics.jsonl, and raw prompts and queries are removed or hashed. Continuous log, trace, and metric exporters are disabled by default.

A network export requires an explicit ctx-telemetry-export command or an operator-enabled exporter configuration. Local JSONL may retain a raw session_id for compatibility, so treat the spool as sensitive. Review the enterprise telemetry guide before enabling export.

ctx-telemetry-export --dry-run --json

The dry run inspects the local spool without exporting it.

CLI Reference

Task	CLI	Guide
Initialize ctx and install graph data	ctx-init	Knowledge graph
Scan a repository and get recommendations	ctx-scan-repo, ctx-recommend	Entity onboarding
Connect an MCP, Python, or CLI host	ctx-mcp-server, ctx	Host integration
Inspect the local runtime	ctx-monitor serve	Dashboard
Review or export telemetry	ctx-telemetry-export, ctx-telemetry-retention	Telemetry

See the full documentation for configuration, APIs, entity lifecycle, and operational details.

Example user stories

Tracker ID	User outcome
CLI-002	Scan a repository and receive a bounded skill, agent, and MCP recommendation set.
CLI-026	Review a custom-model harness recommendation with ctx-harness-install --dry-run before installation.
API-011	Manage local entities through the dashboard's validated API.

Tracking sources

Release readiness is tracked in qa/feature_status.csv. The docs/qa/feature-user-story-status.csv, docs/qa/dashboard-user-story-status.csv, and qa/tool-selection-token-history/tracker.csv files are supporting detail ledgers.

Test Signal

The inventory badge reports pytest collection, not a blanket passing claim. The CI badge links to the change-classified GitHub Actions workflow; individual jobs run the lanes required for a change.

Shipped graph inventory

The shipped artifact contract is a 79,958-node graph covering 68,494 skill entity pages, 467 agents, 10,790 MCP servers, and 207 harnesses.

License

MIT. See LICENSE.