AGENTIC ACCESS Beta

MCP resources and prompts

Canonical safety, deployment and provenance context plus prompts that keep agent output evidence-aware.

Last updated 27 July 2026 · Documentation version 2026.07

Resources

oddessy://docs/overview
oddessy://docs/safety-boundary
oddessy://docs/model-provenance
oddessy://docs/outbound-webhooks
oddessy://docs/monitoring
oddessy://status/hosted
oddessy://fleet/summary

The documentation resources are deterministic summaries with canonical URLs. Hosted status is read live from the same machine-readable snapshot that powers status.oddessy.io.

Resource templates

oddessy://fleet/devices/{device_id}
oddessy://integrations/{integration_id}/events
oddessy://outbound-webhooks/{endpoint_id}/deliveries

Prompts

Prompt	Purpose
`inspect_fleet_health`	Separate observed telemetry from inference, flag stale heartbeats and avoid physical-action claims
`review_model_provenance`	Report model identity, hashes, runtime and verification state without overstating declarations
`triage_active_alerts`	Separate current state from retained transitions and route conclusions to accountable review
`review_webhook_delivery`	Inspect response status, attempts and latency without exposing secrets or payloads

Transport behaviour

The beta is stateless and returns JSON responses to POST requests. GET requests to the MCP endpoint return 405 Method Not Allowed, which is the defined way to indicate that the server does not offer a standalone SSE stream.