IndagateDOCUMENTATION
GET STARTED Live

Start from the skills you already have

Role-based, plain-language routes from first connection to governed industrial AI operation.

Last updated 27 July 2026 · Documentation version 2026.07
No industrial-AI background required

Indagate separates setup into small, testable outcomes. You can start with a browser and a safe sample event, then add edge hardware or plant integrations only when your role requires them.

Five Indagate onboarding pathways for no-code operators, workflow users, software and AI builders, agent users and OT engineers
Every pathway begins with one safe, authenticated outcome and grows through the normal site change-control process.

Choose a pathway

You are coming from...Start withFirst useful outcome
Operations, quality or securityNo-code operatorConnect a workflow and see a verified event without writing code
Zapier, Make, n8n or business systemsAutomation bridgeMap familiar webhook concepts to tenant-scoped industrial events
Data, AI, software or another industryCross-industry bridgeUse MCP to inspect governed fleet context and provenance
OpenClaw or another agent platformAgent pathwayConnect read-only MCP tools and a separate signed event webhook
Controls, robotics or OTIndustrial pathwayEnrol edge compute and define the PLC/safety boundary

Path A: no-code operator

  1. Open Platform → Integrations & agents.
  2. Choose No-code user, then select n8n, Zapier or Make.
  3. Paste the generated HTTPS endpoint and shown-once bearer credential into an HTTP/Webhook step.
  4. Choose Send a safe test event. Confirm the wizard says Verified and a test endpoint appears.
  5. Replace only the sample field mapping with your workflow data. Keep the credential in the platform's secret store.

Definition of done: one authenticated test event appears in your tenant; no production equipment or control path changes state.

Path B: OpenClaw or another LLM agent

  1. Choose AI agent user in Platform.
  2. Create a short-lived personal API token and store it as ODDESSY_API_TOKEN in the agent/Gateway secret environment.
  3. Use the copyable OpenClaw or generic MCP setup and run a connection probe.
  4. Ask the agent to list stale devices or model provenance. It receives read-only tenant context.
  5. Create a separate inbound webhook when the agent workflow must send structured events into Indagate, or an outbound destination when it must receive Indagate lifecycle events.
Three deliberately separate paths

MCP reads governed Indagate context. Inbound webhooks receive events into Indagate. Outbound webhooks deliver signed Indagate lifecycle events to a configured HTTPS receiver. None of these paths writes PLC values, approves deployments or proves a physical action occurred.

Path C: bridging from software, data or robotics

Familiar ideaIndagate equivalent
Workspace or account boundaryTenant partition
Versioned model artifactModel provenance: source, version, runtime and SHA-256
Local service/runnerIndagate Edge runtime on Ubuntu, QCS6490 or Windows
API callbackAuthenticated inbound webhook or signed outbound destination
Agent tool connectorRead-only Model Context Protocol service
Production approval gateHuman, policy and authoritative industrial-control path

Path D: controls and OT engineer

  1. Review the agreed deployment boundary and public safety guidance with the accountable site owner.
  2. Enrol an isolated reference node and verify heartbeat, camera and model manifest.
  3. Connect a non-production event source and validate identity, retry and retention behavior.
  4. Document zones, conduits, credential ownership, acknowledgement paths and fail-safe behavior.
  5. Promote only through the site's existing change-control and risk-assessment process.

When you get stuck

Return to the last verified outcome instead of changing several layers at once. Capture the connector name, timestamp, HTTP status and whether the safe test event appeared. Send that context to info@oddessy.io without including tokens or production data.