# LeanCTX — the Context Gateway for AI systems

> LeanCTX is the Context Gateway for AI systems. It sits between AI and the systems it reads, selects task-relevant context, applies supported controls before delivery, and records what the integration can observe. In context-only integrations, the host owns the model call. The workflow is Select → Control → Learn → Prove.

## Product identity

- Brand: **LeanCTX**, one word. The CLI and package name is `lean-ctx`.
- Product category: **Context Gateway for AI Systems**.
- **Select**, **Control**, **Learn** and **Prove** are the gateway’s four steps.
- LeanCTX by Thinkery AG (CLI and packages: `lean-ctx`, `lean-ctx-bin`, `@thinkery/leanctx-sdk`) is not related to the `leanctx` prompt-compression packages on npm and PyPI.

## What it does

- Retrieves information relevant to a task and prepares it within a context budget.
- Applies PathJail file-access boundaries and configured content filters to supported read results; detected matches can be warned on, redacted or blocked before delivery.
- Reuses useful memory and scoped handoffs across tasks and agents, with source and policy evidence for inspection.
- Can route supported model and tool requests through configured routes, budgets and fallback rules.
- Connects to supported AI tools and can be integrated into products through the SDK and APIs.

## Scope and limits

- Controls apply only to supported calls routed through LeanCTX and depend on the active integration and policy configuration.
- Pattern-based filters can miss sensitive content or prompt injection; they do not provide complete detection.
- Local context processing does not mean local inference. The selected model and deployment determine where model requests are processed.
- Recorded usage, estimated costs and task outcomes depend on the workflow; review the measurement scope before comparing results.

## Offer boundaries

- Community Engine: open-source local context processing with supported tools. Local processing does not imply local inference; the chosen model and deployment determine where model requests run.
- Enterprise: a commercial, release-based offer for organizational controls and deployment. Confirm the supported features in the named licensed release and written agreement.
- SDK / OEM: five stable SDK primitives and six Agent Tools language contracts. Preview operations remain under the Python namespace `leanctx_sdk.preview`; production and OEM rights require a separate signed written agreement.

## LeanCTX Guided Onboarding

- English: https://leanctx.com/onboarding/
- Deutsch: https://leanctx.com/de/onboarding/
- One service combining workflow assessment, guided setup, hands-on team education, joint evaluation and handover.
- One agreed workflow in one test environment through an existing supported integration and model path.
- Deliverables: a test configuration, practical team training, written findings, workflow documentation and a recommendation for further use.
- Before commissioning, agree in writing the workflow, approved sources and test data, setup, participants, sessions, schedule, evaluation criteria and customer involvement.
- Custom connectors, additional workflows, production rollout and ongoing operations require a separate scope. Existing subscriptions, model usage, infrastructure and additional software licences are separate.
- Community remains available for independent installation and use under its licence. Annual Enterprise contracts are quoted separately for organizational capabilities, deployment scope and support; expansion to additional teams, workflows and environments is agreed as needed.
- Booking the introductory Microsoft Teams call does not commission a paid engagement. No automatic subscription or follow-on purchase.
- Evidence covers only the selected supported test path and enabled rules. Delivery evidence does not establish later model use; the host retains model and task choice.
- Contact: hello@leanctx.com.
## Current references

- [Use cases](https://leanctx.com/use-cases/): twelve everyday applications across six industries, explained for business leaders. Teams can review contract and software changes, investigate failed tests, compare experiments and pass on service knowledge. Each scenario explains the problem, relevant information, LeanCTX's contribution and what the team can review or decide. These are illustrative applications, not customer references.

- [Product overview](https://leanctx.com/)
- [What is LeanCTX](https://leanctx.com/what-is-leanctx/)
- [How it works](https://leanctx.com/how-it-works/)
- [Deployment options](https://leanctx.com/enterprise/)
- [Pricing](https://leanctx.com/pricing/)
- [Documentation](https://leanctx.com/docs/)
- [Getting started](https://leanctx.com/docs/getting-started/)
- [Architecture](https://leanctx.com/docs/architecture/)
- [SDK](https://leanctx.com/docs/sdk/)
- [API reference](https://leanctx.com/docs/api-reference/)
- [MCP tools and schemas](https://leanctx.com/docs/tools/)
- [Security](https://leanctx.com/docs/security/)
- [Changelog](https://leanctx.com/docs/changelog/)

## Machine-readable resources

- [MCP tool schemas](https://leanctx.com/generated/mcp-tools.json)
- [Sitemap index](https://leanctx.com/sitemap-index.xml)
- [Repository](https://github.com/yvgude/lean-ctx)

These files are optional reading aids. Google says its Search systems, including generative AI search features, do not use `llms.txt`; publishing a summary does not guarantee crawling, indexing or citation. See [Google's guidance](https://developers.google.com/search/docs/fundamentals/ai-optimization-guide).