Cleric

Learning Objectives

• Describe what Cleric does and how a read-only AI SRE teammate approaches incidents
• Explain how hypothesis-driven reasoning and transparent hypothesis trees work
• Identify why a safety-first, human-in-the-loop design is the product's differentiator

What Is Cleric?

Cleric is an AI site-reliability engineering (SRE) teammate that helps on-call engineers investigate alerts. When an alert fires, Cleric investigates using hypothesis-driven reasoning: instead of pattern-matching blindly, it forms candidate explanations for what might be wrong, gathers evidence, and tests those hypotheses much the way a thoughtful human engineer would. Founded in 2023 and based in San Francisco, Cleric is a private, early-stage company.

What most distinguishes Cleric is its safety posture. It shows its work as a transparent hypothesis tree — a visible structure of the possibilities it considered and the evidence for and against each — so engineers can follow and trust its reasoning. And it stays strictly read-only until a human approves any remediation. Cleric investigates autonomously but acts only with human sign-off, making the "investigate freely, change nothing without approval" boundary its core design principle.

What Cleric Does

• Alert investigation — works on-call alerts by investigating what triggered them
• Hypothesis-driven reasoning — forms candidate explanations and tests them against evidence, rather than guessing
• Transparent hypothesis tree — shows the possibilities considered and the supporting or contradicting evidence for each
• Read-only by default — investigates without changing anything until a human approves
• Layered memory — draws on semantic, episodic, and procedural memory to reason about systems, past incidents, and how-to knowledge
• Human-approved remediation — any fix requires explicit human sign-off before it is applied

How AI Is Applied

Cleric's AI is organized around structured reasoning rather than a single opaque answer. Its hypothesis-driven approach mirrors how a strong SRE thinks: given an alert, what are the plausible causes, and what evidence would confirm or rule each one out? It draws on three kinds of memory to do this well — semantic memory (general knowledge about systems and technologies), episodic memory (what happened in past incidents), and procedural memory (the steps and playbooks for how to investigate and resolve issues). Combining these lets it reason with context rather than starting from scratch each time.

The transparent hypothesis tree is what makes that reasoning trustworthy. Because engineers can see the branches Cleric explored and the evidence behind them, they can judge whether the conclusion is sound instead of taking it on faith. Paired with the strict read-only stance until human approval, this is a deliberately safety-first design: autonomy in investigation, human control over action. For an early-stage product operating in the sensitive territory of production incidents, that "investigate autonomously, act only with approval" boundary is both the honest limit and the genuine differentiator.

Who Uses Cleric

Cleric is aimed at site-reliability engineers, DevOps teams, and on-call engineers who want AI help investigating alerts without ceding control over changes to production. It particularly suits teams that value transparency and safety — those who want to see an agent's reasoning and keep a human hand on any remediation — and organizations comfortable adopting an early-stage tool in exchange for its safety-first design.

Pricing

Cleric is an enterprise product, and as an early-stage company its pricing is arranged directly with customers based on scope and usage. Specific terms are not published, so organizations contact Cleric directly to discuss access and pricing. Prospective users should expect the engagement typical of an early-stage vendor rather than self-serve, published tiers.

Company Details

Strengths

• Safety-first by design — read-only until a human approves any remediation, so it cannot break production on its own
• Transparent reasoning — the hypothesis tree lets engineers see and trust how it reached a conclusion
• Structured investigation — hypothesis-driven reasoning mirrors how a strong human SRE works
• Context-aware memory — semantic, episodic, and procedural memory ground its reasoning in real knowledge and past incidents
• Human control preserved — pairs automated investigation with human judgment over every fix

Limitations and Considerations

• Early-stage — a young, small company; maturity and scale are still being proven
• Investigation, not action — by design it does not remediate on its own, so humans remain in the resolution loop
• Direct-engagement pricing — no published tiers; access is arranged with the vendor
• Results depend on data and integrations — reasoning quality relies on the evidence it can gather from a team's systems

Key Takeaways

• Cleric is an AI SRE teammate that investigates on-call alerts with hypothesis-driven reasoning and shows its work as a transparent hypothesis tree
• It stays read-only until a human approves any remediation, making "investigate autonomously, act only with approval" its core, safety-first differentiator
• Semantic, episodic, and procedural memory ground its reasoning, though it is an early-stage product still proving out at scale
• Best for SRE and DevOps teams that want transparent, safe AI-assisted investigation while keeping humans in control of every production change