AI SRE platform that instruments services, monitors production, verifies pull requests after deployment and drives incidents toward proposed fixes.
Product sources: OnePatch official product/pricing surface + Y Combinator F26 company page
OnePatch is an AI SRE platform built around production telemetry and the software-delivery loop. It can ingest OpenTelemetry logs, metrics and traces, instrument repositories, build monitors and dashboards, and use production signals to investigate incidents and verify pull requests after deployment.12 Investigations can be launched from the product chat, Slack, MCP or CLI, while GitHub integration ties operational findings back to source code and deployed changes.2
The product surface is more focused than a general observability suite: its differentiator is joining telemetry, repository context and AI-led incident investigation into one workflow. The current site also positions OnePatch around proposing fixes and driving incidents toward resolution, but AiToolMap did not independently test remediation quality or autonomous action behavior.1
Current pricing is enterprise-oriented. Starter costs $500/month and includes 25 conclusive investigations, with additional investigations at $20 each. Scale costs $2,000/month and includes 100, with $10 additional investigations; Enterprise pricing is custom.1 There is no public free tier in the current catalog.
The outcome-based investigation allowance makes the commercial model easier to map to incident volume than raw token or seat usage, but the entry price can be high for small teams or low-incident services. Buyers should estimate expected investigation volume and confirm what counts as a “conclusive investigation” in their contract before forecasting overages.
OnePatch's Privacy Policy, last updated September 1, 2026, describes the company as a processor for customer telemetry and lists AWS, Anthropic, OpenAI, ClickHouse Cloud, WorkOS and Stripe among subprocessors.3 It states that AI subprocessors are contractually restricted from training on API-submitted customer data and describes encryption, access control, audit and other security measures.3
The Terms, effective July 1, 2026, state that Customer Data remains owned by the customer and is processed to provide and improve the service; they also make clear that operational judgment and production responsibility remain with the customer.4 Those are useful contractual and transparency signals, but they remain first-party claims rather than an independent security audit or reliability benchmark.
The fixed independent review panel produced no qualifying exact-product evidence for OnePatch. That absence is treated neutrally under Rating v1.0 rather than as a negative score. Y Combinator independently corroborates the active Fall 2026 company and AI-SRE product surface, but does not provide a reliability or performance benchmark.5
OnePatch is best suited to software engineering, SRE and DevOps teams already operating production services with GitHub and OpenTelemetry-based observability. It is particularly relevant when teams want incident investigation, deployment verification and code context to live in the same AI-assisted operational loop.12
It is less suitable for very small teams looking for a free or low-cost monitoring product, organizations that require independent security certification before adoption, or buyers that only need dashboards without AI-led investigation. Production teams should retain human operational ownership and validate proposed fixes through their own change-control process.4
coherent AI-SRE workflow from instrumentation and monitoring through incident investigation and post-deploy verification; GitHub, Slack and OpenTelemetry integration; multiple control surfaces through chat, CLI and MCP; transparent current pricing; and unusually specific privacy/subprocessor language for a young product.123
no free tier; $500/month entry pricing; very young product; no eligible independent product benchmark or user-rating evidence in the canonical panel; and vendor-authored reliability, security and remediation claims that should not be treated as independently proven.
Products matched by shared tasks, audience, workflow, product type and capabilities.