Quickstart

1. Run the proof

From the project you want to govern:

runeward quickstart

It creates .runeward/quickstart.toml (an existing policy isn't overwritten), runs an allowed command, blocks a recursive delete and verifies the signed audit trail. --no-run only creates the policy.

If it fails, runeward doctor quickstart (or --json) checks policy, runtime, image and state, and suggests fixes.

2. Run your own command

runeward --config-dir .runeward print quickstart
runeward --config-dir .runeward enter quickstart -- uname -a

quickstart is the policy name.

3. Open the dashboard

Give each running instance its own state directory, because the audit ledger is single-writer:

RUNEWARD_STATE_DIR=/tmp/runeward-quickstart \
  runeward --config-dir .runeward serve

At http://localhost:8080, drive sandboxes, resolve approvals, inspect policy, network, budget and audit events, download the workspace or signed evidence, and create and restore snapshots.

4. Verify the signed audit trail

Stop the server first, since it holds the same ledger:

runeward --config-dir .runeward evidence export quickstart -o runeward-evidence.json
runeward evidence verify runeward-evidence.json
runeward policy learn runeward-evidence.json > proposed-policy.toml

policy learn prints proposals for human review. It doesn't change the policy. For what each check covers, see Overview.

5. Work on your own code

Rune Box copies your project once into an isolated /workspace. It doesn't mount the host folder. Set this in .runeward/quickstart.toml, or in the dashboard's create dialog, before launching:

[host]
copy_from = "/absolute/path/to/your-project"

runeward --config-dir .runeward enter quickstart --keep
runeward export <sandbox-id> ./agent-output

Next: Policies and Security.