Sheaf Pod CLI

Call a pod of AI minds — one reconciled answer plus a coherence audit — from the command line or an automated agent pipeline. Thin wrapper over the Sheaf Pod API. No dependencies; Node ≥ 18.

Install

No package manager needed — it's a single zero-dependency file (Node ≥ 18):

# fetch it
curl -O https://sheaf.one/sheaf-pod.mjs
# run it
node sheaf-pod.mjs ...
# or make it executable
chmod +x sheaf-pod.mjs && ./sheaf-pod.mjs ...

Setup

export SHEAF_POD_KEY=sk_pod_your_key      # required
export SHEAF_POD_URL=https://api.discursa.ai   # optional (default)

Use

# Run a pod, get the answer on stdout (pipe it into your next step)
sheaf-pod run -f examples/marketing-video.json -i "Brief: promote MyPeopleDatabase to founders"

# Input from stdin
echo "the brief" | sheaf-pod run -f examples/marketing-video.json

# Headless / nightly: submit + poll until done (best for long runs), or use a webhook
sheaf-pod run -f pod.json -i "..." --async
sheaf-pod run -f pod.json -i "..." --async --webhook https://your-app.com/cb

# Live progress (each mind as it lands) to stderr, answer to stdout
sheaf-pod run -f pod.json -i "..." --stream

# Full JSON instead of just the answer
sheaf-pod run -f pod.json -i "..." --json

# Saved pods + account
sheaf-pod save -f examples/marketing-video.json --name marketing-video
sheaf-pod list
sheaf-pod run --pod-id <id> -i "..."
sheaf-pod usage

In an automated pipeline

stdout is just the answer (or JSON with --json), stderr is progress, exit code is non-zero on failure — so it composes cleanly:

BRIEF=$(your-agents-write-the-brief)
PACKAGE=$(sheaf-pod run -f marketing-video.json -i "$BRIEF" --async --json)
echo "$PACKAGE" | your-renderer-and-youtube-uploader

See the full API + pod schema reference at https://sheaf.one/pod-integration.md