Handbook
LCDL hero and messaging concept
Kicker: Governed LLM tasks for Python systems
Updated
Homepage hero candidate
Kicker: Governed LLM tasks for Python systems
Headline: Build reliable LLM workflows without scattering prompts, transport quirks, or JSON recovery code.
Subhead: Forge LCDL gives teams versioned task contracts, schema-aware outputs, deterministic control-flow operators, and OpenAI-compatible transport helpers—so LLM calls can be tested, reviewed, and operated like governed software components.
CTAs
- Run your first governed task
- Browse the task catalog
Proof pills
- Offline-first examples
- Contract + schema sidecars
- Agent/Cursor ready
Section messages
| Section | Message |
|---|---|
| What LCDL is | A governed task layer for synchronous LLM calls. |
| What it replaces | Scattered prompts, ad-hoc retries, fragile JSON parsing, undocumented outputs. |
| How it works | Payload → task contract → runner → transport → parse/validate → Ok/Err. |
| Who uses it | Developers, task authors, operators, architects, agents. |
| Enterprise boundary | LCDL provides primitives; host systems own credentials, deployment, data, and SLAs. |