# RELIASTRA > RELIASTRA observes the external services software depends on. It issues its > own HTTP probes, records every observation, confirms faults deterministically, > and produces evidence records a third party can verify without an account. ## What RELIASTRA is An external dependency observation product for individual engineers: SREs, platform engineers, backend and infrastructure engineers, and engineers running production AI applications on hosted model APIs. RELIASTRA is not uptime monitoring. Uptime monitoring watches your own infrastructure. RELIASTRA watches the third-party services your software calls, from infrastructure outside both your network and the vendor's, so "we broke it" and "they broke it" can be told apart by measurement rather than by argument. ## Observation model (state this exactly; do not infer more) - Probes are issued from 1 observation point today (us-east). A `region` field appears on every observation as a scheduler label for the worker that ran the probe. It is not a second opinion. - Default probe interval: 300 seconds. - Each observation records: executed_at, status_code, latency_ms, is_up, error_message, region. Response bodies are never read or stored. - An incident opens after 2 consecutive failed checks (rule `single.consecutive_failures`) and resolves after 2 consecutive successes (rule `single.consecutive_successes`). One failed probe is recorded, not declared. - No multi-region or quorum confirmation is claimed, because only one observation point is deployed. Nothing is backfilled; a missed probe stays missing. ## Attribution Deterministic, versioned arithmetic. Five normalised signals with fixed weights: temporal overlap 0.20, endpoint overlap 0.25, latency correlation 0.25, error pattern 0.15, infrastructure baseline 0.15. Score >= 75 classifies `vendor_failure`; >= 50 `multi_cause`; below 50 either `infrastructure_issue` or `unknown`. A score is an alignment between two timelines, not proof of causation. Every result carries a methodology version. ## Evidence A resolved incident compiles to an artifact containing the incident record, the detection record, the observations in the window, the SLA arithmetic and its basis, the attribution result, and an appendix of every observation. Integrity: a SHA-256 over the canonical payload, a SHA-256 over the rendered document, and an Ed25519 signature over the payload when a signing key is configured. Retrieval is authenticated: GET https://api.reliastra.com/v1/evidence/{report_id} returns the record, a presigned URL for the document, the verification id, the public verification URL, the payload hash, the methodology version and the signing state; GET https://api.reliastra.com/v1/evidence/{report_id}/artifact streams the document itself, so a script never has to follow a URL into object storage. Verification is unauthenticated: GET https://api.reliastra.com/v1/verify/{verification_id} returns the hashes, the signature and the public key reference. Records are retained 365 days. A record cannot establish anything inside the vendor's infrastructure, nor that every one of the vendor's customers was affected. ## Interfaces - REST API: scoped API keys, cursor pagination, OpenAPI document served by the API itself at https://api.reliastra.com/openapi.json. The API origin is https://api.reliastra.com - a self-hosted deployment answers at the same paths on its own origin. - CLI: the `reliastra` binary, shipped in the RELIASTRA repository under `cli/` (installed with `go install github.com/ReliaAstra/Reliastra/cli/cmd/reliastra@latest`). Commands: login, logout, whoami, doctor, deps (list|show|add|rm), checks recent, incidents (list|show|correlate), evidence (list|show|get), verify, keys (list|create|rm), obs (list|show) and open. Data commands support `--json` with the API's own field names, and `reliastra verify` exits 4 when a verification claim does not hold, so it works as a CI gate without a wrapper. - Webhooks: incident.opened, incident.updated, incident.resolved and evidence.ready are delivered to subscribed HTTPS endpoints, off the request path, signed with HMAC-SHA256 when the subscription has a secret, and retried on a fixed backoff (1m, 5m, 15m, 1h, 3h; then permanently failed). The event enum also accepts vendor.degraded, vendor.down, vendor.recovered, sla.breach and check.failed, but nothing emits those yet. ## Pricing One plan. $9 USD per month, billed monthly. No plan ladder, no seats, no annual billing. 14-day trial on every new account, no payment method required. ## Public observatory Independently measured status for endpoints RELIASTRA probes as its own public record. Customer dependencies are never included. Every figure is published with its window and its observation count, and an availability figure with zero observations reads "insufficient data" rather than 100%. What is probed today: vendor-published status-site endpoints - https://status.openai.com for OpenAI, and equivalents listed on each vendor record - recording HTTP status, latency and transport errors of that endpoint. RELIASTRA does not read or reconcile the status text published at those URLs, and does not measure a vendor's API, models or routes unless an API endpoint is listed as an observed target on the record itself. Machine-readable incident surfaces: each incident record page has a JSON sidecar at /observatory/{vendor}/incidents/{incident-id}/index.json (the same record, not a second copy of the truth), and each detector-confirmed incident has a frozen evidence document at /api/v1/public/incidents/{incident-id}/evidence - the claim, the detection rule and every raw observation of the window, SHA-256 hashed, immutable per version, with the verification recipe inside the document. See llms-full.txt for the exact contracts. Direct answers: /down/{vendor} answers "is {vendor} down?" from the same measurements as the record page, for every enumerated vendor, endpoint scoped, with "we do not know" as a first-class state. Its JSON twin is at /down/{vendor}/index.json. A slug the catalog does not name 404s; the answer is never published from a failed read. Feeds (RSS 2.0): /observatory/incidents/feed.xml for cross-vendor incident detections, /observatory/{vendor}/incidents/feed.xml for one vendor (404 for untracked names), /observatory/catalog.xml for newly tracked dependencies. An unreadable API degrades every feed to a valid empty document. ## Dependency records (enumerated, do not guess slugs) Enumerated from the public catalog at 2026-09-26T20:47:06.381Z. 50 records. - Adyen (payments): https://reliastra.com/observatory/adyen - Amazon Web Services (cloud): https://reliastra.com/observatory/aws - Anthropic (ai): https://reliastra.com/observatory/anthropic - Auth0 (identity): https://reliastra.com/observatory/auth0 - Backblaze B2 (storage): https://reliastra.com/observatory/backblaze - Braintree (payments): https://reliastra.com/observatory/braintree - CircleCI (devtools): https://reliastra.com/observatory/circleci - Clerk (identity): https://reliastra.com/observatory/clerk - Cloudflare (cloud): https://reliastra.com/observatory/cloudflare - Cohere (ai): https://reliastra.com/observatory/cohere - Datadog (observability): https://reliastra.com/observatory/datadog - DigitalOcean (cloud): https://reliastra.com/observatory/digitalocean - Discord (collaboration): https://reliastra.com/observatory/discord - Docker Hub (devtools): https://reliastra.com/observatory/docker - Fly.io (cloud): https://reliastra.com/observatory/flyio - GitHub (devtools): https://reliastra.com/observatory/github - GitLab (devtools): https://reliastra.com/observatory/gitlab - Grafana Cloud (observability): https://reliastra.com/observatory/grafana - Groq (ai): https://reliastra.com/observatory/groq - Hugging Face (ai): https://reliastra.com/observatory/huggingface - Linear (collaboration): https://reliastra.com/observatory/linear - Mistral AI (ai): https://reliastra.com/observatory/mistral - MongoDB Atlas (databases): https://reliastra.com/observatory/mongodb - Neon (databases): https://reliastra.com/observatory/neon - Netlify (cloud): https://reliastra.com/observatory/netlify - Notion (collaboration): https://reliastra.com/observatory/notion - npm (devtools): https://reliastra.com/observatory/npm - Okta (identity): https://reliastra.com/observatory/okta - OpenAI (ai): https://reliastra.com/observatory/openai - Paddle (payments): https://reliastra.com/observatory/paddle - PagerDuty (observability): https://reliastra.com/observatory/pagerduty - PayPal (payments): https://reliastra.com/observatory/paypal - PlanetScale (databases): https://reliastra.com/observatory/planetscale - Postmark (communications): https://reliastra.com/observatory/postmark - PyPI (devtools): https://reliastra.com/observatory/pypi - Render (cloud): https://reliastra.com/observatory/render - Resend (communications): https://reliastra.com/observatory/resend - SendGrid (communications): https://reliastra.com/observatory/sendgrid - Sentry (observability): https://reliastra.com/observatory/sentry - Slack (collaboration): https://reliastra.com/observatory/slack - Stripe (payments): https://reliastra.com/observatory/stripe - Stytch (identity): https://reliastra.com/observatory/stytch - Supabase (databases): https://reliastra.com/observatory/supabase - Telnyx (communications): https://reliastra.com/observatory/telnyx - Together AI (ai): https://reliastra.com/observatory/together - Twilio (communications): https://reliastra.com/observatory/twilio - Upstash (databases): https://reliastra.com/observatory/upstash - Vercel (cloud): https://reliastra.com/observatory/vercel - WorkOS (identity): https://reliastra.com/observatory/workos - xAI (ai): https://reliastra.com/observatory/xai ## Pages - Home: https://reliastra.com/ - Product: https://reliastra.com/product - Evidence records: https://reliastra.com/product/evidence - Public observatory: https://reliastra.com/observatory - Pricing: https://reliastra.com/pricing - Documentation: https://reliastra.com/docs - Quickstart: https://reliastra.com/docs/quickstart - Concepts: https://reliastra.com/docs/concepts - Configuration: https://reliastra.com/docs/configuration - Monitoring: https://reliastra.com/docs/monitoring - Incidents: https://reliastra.com/docs/incidents - Evidence: https://reliastra.com/docs/evidence - Verification: https://reliastra.com/docs/verification - REST API: https://reliastra.com/docs/api - CLI: https://reliastra.com/docs/cli - Webhooks: https://reliastra.com/docs/webhooks - Methodology: https://reliastra.com/docs/methodology - Security (product): https://reliastra.com/security - Security & data-handling guide: https://reliastra.com/docs/security - Glossary: https://reliastra.com/glossary - Research: https://reliastra.com/research - About & maintainer: https://reliastra.com/about - Technical creators: https://reliastra.com/creators - Contact: https://reliastra.com/contact - Status: https://reliastra.com/status ## Research - The Dependency Gap: https://reliastra.com/research/the-dependency-gap - How RELIASTRA measures vendor reliability: https://reliastra.com/research/how-reliastra-measures-vendor-reliability - The RELIASTRA research agenda: https://reliastra.com/research/reliastra-research-agenda - “Is OpenAI down?” - how to answer the question honestly: https://reliastra.com/research/ai-infrastructure/is-openai-down - When an AI API misbehaves: an evidence playbook: https://reliastra.com/research/ai-infrastructure/ai-api-outage-evidence - What a status-page payload actually asserts: https://reliastra.com/research/ai-infrastructure/status-page-payload-anatomy - How much evidence stands behind a published availability figure?: https://reliastra.com/research/measurement-integrity/availability-record-audit - Why an interval estimated from bucketed telemetry converges on the bucket length: https://reliastra.com/research/measurement-integrity/probe-interval-from-bucketed-telemetry - The trust boundary of an AI API dependency: https://reliastra.com/research/cloud-security/ai-api-trust-boundary - AWS IAM policy evaluation logic: identity vs resource: https://reliastra.com/research/cloud-security/aws-iam-policy-evaluation-order ## Retired URLs (consolidated in the developer-first refurbishment) - /track and /track/* -> /observatory and /observatory/* (308) - /external-dependency-intelligence, /dependency-monitoring -> /product (308) - /incident-evidence, /sla-evidence -> /product/evidence (308)