Headless and piping

Seepient functions as a standard Unix command-line utility. It reads input from standard input (stdin), writes results to standard output (stdout), and prints errors and progress status to standard error (stderr).

---

Piping data via stdin

You can pipe data directly from other command-line tools into Seepient:

Reviewing git diffs

git diff main...feature-branch | seepient "Review this diff for missing test cases and edge cases"

Analyzing logs

tail -n 200 /var/log/app.log | seepient "Identify unique error spikes and summarize probable causes"

Parsing test failure outputs

pnpm test 2>&1 | seepient -y "Fix the broken unit tests identified in the test output"

When stdin is attached, Seepient combines the piped stream with your prompt argument before querying the language model.

---

Continuous integration pipelines

Seepient integrates cleanly into CI runners (such as GitHub Actions, GitLab CI, or Jenkins).

Example GitHub Action step:

- name: Run Seepient Code Audit
  env:
    ANTHROPIC_API_KEY: ${{ secrets.ANTHROPIC_API_KEY }}
    SEEPIENT_CONSENT_MODE: autonomous-trusted
  run: |
    seepient -y "Inspect modified files in this pull request and check compliance with project guidelines"

Because stdout and stderr are separated:

• Agent responses and generated summaries write to stdout (safe for piping to PR comments or build artifacts).
• Sandbox notifications, status spinners, and diagnostic logs write to stderr.

---

Multi-instance swarm loops

Because Seepient maintains a lightweight memory footprint and low startup overhead, you can orchestrate multi-agent batch tasks using standard shell loops or xargs:

#!/usr/bin/env bash
# Run parallel security reviews across all microservice folders

for dir in services/*; do
  (
    cd "$dir" || exit 1
    echo "Auditing $dir..."
    seepient -y "Run a security audit on this service and write findings to SECURITY_REPORT.md"
  ) &
done

wait
echo "All audits completed."