Documentation
DeepClause turns specs and SOPs into executable DML harnesses. The logic that must hold is code; the model drafts and explains. Every change, run, and artifact is reviewable.
DeepClause has two main functions: sop2harness, the CLI that turns SOPs into a versioned harness, and the pi extension, which authors and runs DML inside a pi session. Both share one DML runtime and one harness format.
s2h scaffolds a project, authors one validated DML
Skill per procedure through a pi session, versions the harness, and
exports it as a self-hosted API, web chat, and MCP server.
s2h create authors Skills from SOPss2h commit validates, versions, and tagss2h export produces API + chat + MCP + DockerThe pi extension turns a coding session into a DML authoring and execution environment. It ships two workflows, both producing DML for the same runtime.
Both share one DML runtime: one model, one audit trail. See the DML language reference and the DML SDK/CLI.
Not using pi or sop2harness? Run and embed DML directly with the DML SDK & CLI guide. For calibrated semantic decisions, see decision models.
One CLI from SOP to a versioned harness — then embed the runtime anywhere.
npm install -g deepclause-sop2harness
s2h init
s2h create "Turn this SOP into a harness"
s2h commit -m "add triage workflow"
s2h export --out ./exportScaffold, author through a pi session, version, and export a self-hosted API, web chat, and MCP server.
pi install \
git:github.com/deepclause/deepclause-piAuthor and run DML in your session with SDD (SpecKit) and the Handbook → DML skill. Uses pi's model and credentials.
import { createDeepClause } from "deepclause-sdk";
const dc = await createDeepClause({
model: "host-active-model",
llmBackend, // your model
});Embed the DML runtime in your own service with a host-provided model backend; no provider keys required.
s2h CLI, or the SDK's injectable model backend. The runtime tool
whitelist stays explicit.
Sop2Harness CLI reference → A concrete sop2harness example →