原始内容
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 Python 3.11 or newer.
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 six project-owned, MIT-licensed, no-key
fallbacks: ctx-python-testing, 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.