Licence
MIT
Version
0.1.43
Deps
3
Size
27 kB
Vulns
0
Weekly
6.4K
@agentic-surfaces/cli
The command-line runner for agentic-surfaces —
a local-first, agent-in-the-loop workflow engine. Loads a project's agentic-surfaces.config.yaml
workflows/*.yaml+agents/*.md, runs the scheduler, and serves the live dashboard.
Repo & full docs: https://github.com/dalebaldwin/agentic-surfaces
Usage
npx -y @agentic-surfaces/cli@latest <command> [projectDir] [--ui] [--watch] [--fake]| Command | What it does |
|---|---|
start <dir> |
Run the scheduler (cron triggers fire). --ui adds the dashboard; --watch restarts on config edits. |
run-once <workflow.yaml> |
Run one entry workflow once, ignoring its cron. |
ui [dir] |
Dashboard + scheduler (the default when no command is given). |
validate <workflow.yaml> |
Parse-check a workflow. |
Pin a version for reproducibility: @agentic-surfaces/cli@0.1.x.
Secrets
Workflows use ${VAR}. Supply via a .env in the working directory (auto-loaded), a secrets
manager like Doppler, or plain exports. Real env / Doppler override .env.
Dashboard
--ui serves http://127.0.0.1:4000 — workflow graph (with the taken path highlighted), live runs,
per-node output, agent activity (reasoning + tool/MCP calls), and a cache inspector. Set
FLOW_UI_PORT to change the port, FLOW_NO_OPEN=1 to not open a browser.
MIT licensed.