Developer docs

Hassle is a local-first CLI. Run it yourself or hand it to your agent — Claude, Codex, Gemini or xAI. Everything that touches your code runs on your machine; Hassle stores only derived, signed evidence.

Install & login

Install the CLI and bind your workspace with a device code. Login happens where your company is built so Hassle can attach to your local scanner.

$ curl -fsSL hassle.dev/install | sh
$ hassle login --device
  visit hassle.dev/device · code HX-92KQ
  ✓ workspace bound · evidence room ready

This command requires an approved beta account.

Using Claude Code, Codex or Cursor?

Paste Hassle's guided setup prompt into your coding agent. The agent runs the CLI workflow from your repo, explains each step, and asks before publishing evidence.

You don't need to memorise terminal commands. Start in the founder portal, copy the full prompt, then paste it into Claude Code, Codex or Cursor with your repo open. The agent walks the workflow; you confirm each publish.

You are helping me set up Hassle for my company proof room.

Goal:
Run the Hassle local proof workflow from this repository so I can create source-backed claims for investors without uploading raw code by default.

Rules:
- Do not expose secrets, .env files, private keys, credentials or customer data.
- Use preview commands before publishing anything.
- Before running any publish command, explain what will be shared and ask me for confirmation.
- Explain each step in plain English.

# … full prompt continues with the hassle login → init → scan → publish sequence.

Get the full prompt in your founder portal →

Scan & route

A scan runs deterministically and locally, then optionally generates a model summary. Verified claims never depend on the model.

$ hassle scan
  → Running deterministic local scan
  ✓ framework detection
  ✓ dependency graph
  ✓ API route analysis
  ✓ secret scan
  ✓ test maturity
  ✓ CI/CD detection

  Model summary:
  Using Hassle managed model gateway

  Run hassle model configure to use your provider or local/private mode.

Model modes

Hassle separates proof from narrative. The scanner produces deterministic, source-backed findings. A model then summarises and explains them. You choose which model does the summarising — it never changes verification status.

A fifth mode — no-model deterministic scan — skips summarisation entirely and emits only structured findings.

Preview exactly what is sent

The most important command for trust. Inspect the model payload before any call is made.

The model will receive:
✓ structured architecture findings
✓ dependency summary
✓ test maturity summary
✓ security finding categories

The model will not receive:
✗ raw source files
✗ .env files
✗ secrets
✗ customer data
✗ raw logs

Model commands

Local / private mode

Advanced founders can configure local models for private summarisation. Hassle still relies on deterministic scanner output, source-backed claims, schema validation and signed evidence reports. Local model summaries are labelled separately from verified claims.

$ hassle model configure local --provider ollama --model llama3.1
  ✓ local model configured

$ hassle scan --model local
  Model summary:
  Using local model: Ollama llama3.1
  Payload: structured findings only
  Raw code sent: no
  Note: local model summaries are labelled separately from verified claims.

Post-raise: operate

When the round closes, the operating commands prepare board-ready truth from connected runway, GTM and product proof. The CLI prepares operating truth; the dashboard publishes it.

Post-raise: board & meetings

Post-raise: actions, updates & follow-on