TrustOps onboarding · Replay-first

Run your first replay.

Your Trust Agent guides inspection and evidence review. This path mirrors how Recon Agent · Starter onboards you: local trust receipts, seeded guided replay, then GhostLog and timeline surfaces—inspect-only posture, no autonomous actions. Aliethia still handles first-contact framing when you open guided mission flows.

Recon Trust Agent · Starter posture

This is the detailed version of the 5-minute arc: keep your runtime, install, run replay, open GhostLog / timeline, then read the Trust Agent posture summary—without taking autonomous action. This page renders the starter spine as checklist links (not a live chat panel yet).

Local starter mode · seeded vs live telemetry

Commands below write local trust receipts under .reconai / .recon — no cloud required for the CLI vignette. Web surfaces may show empty, demo-seeded, or org-backed rows depending on your deployment; production fleet telemetry requires org ingestion wiring.

Step A — Install

npm install @reconai/sdk

If install fails with workspace:, the tarball on npm predates dependency fixes—you need @reconai/sdk@^0.1.7 (or newer) published, or wire the CLI from source. Contributing in recon-ai-trustops: skip global npm install; use pnpm reconai:demo from the repo root (same demo as npx; see Step B note).

LangChain adapter (Python): reconai-langchain (install from GitHub until PyPI publishes; see repo README).

Step B — Run (local replay-oriented receipts)

npx reconai demo
  • · Writes rolling reflex benchmark lines under your cwd (starter workspace).
  • · Seeds material you can correlate with dashboards when ingestion is wired.
  • · Stays offline in default dry_run paths.

recon-ai-trustops checkout: run pnpm reconai:demo from the workspace root, not pnpm exec reconai demo from a nested directory—pnpm may not expose the CLI on Windows there. Receipts write under cwd; run from whichever folder should own .recon/.

Step C — Inspect in the Trust OS (after CLI)

Open product surfaces in any order. Prefer guided replay first if you want a deterministic seeded walkthrough before your own trace volume exists.

Timeline / GhostLog often stay sparse until ingestion is connected — that is expected for a fresh browser session. Guided replay stays the fastest deterministic “ontology” preview.

What you're seeing (starter vocabulary)

  • Trust State

    Trust states are individual trust evaluations; today’s count shows how actively your systems are being judged and governed in real time.

  • Governance Intervention Rate (GIR)

    GIR shows how often governance has to intervene—approvals, blocks, or policy holds—indicating how much human or policy load the fleet is creating.

  • Drift signal

    Drift signals flag behavioral change—like reliability or intervention trends moving the wrong way—before they harden into customer-impacting incidents.

  • Replay & chronology

    The live timeline tells the story of your fleet’s day—what executed, what was held, and where trust broke—in one auditable stream.

  • Autonomy posture

    Autonomy status shows the real operating posture of your stack—who runs free, who is watched, and who is constrained by policy.

Copy uses the demo lens from the Trust Literacy layer—plain-language posture, not bespoke marketing claims.

Agent-suggested next steps

Same destinations as before—framed as what Recon Agent · Starter would queue after your first replay.

Connect LangChain

Python adapter + runnable quickstart traces.

Adapter repo →
Production ingestion

Point guard() / ingest at your Recon deployment.

API overview →
Org mission

Aliethia first contact + starter caps.

Guided mission →
Starter runtime missions

ReflexFlow trust-runtime onboarding — three governed missions with preview receipts.

View starter missions →
Live timeline

Org-scoped when trust states are flowing.

Open timeline →