Cleric is an AI SRE agent from Agentik, Inc. that follows code changes into production. It detects regressions, investigates the cause, and opens a fix pull request for human review.
It works mainly in Slack and a web app. Cleric reads release messages, splits each release into changes (usually one per pull request), and checks each change against real traffic.
Cleric also handles alerts. It acknowledges pages, closes false positives, and groups duplicate low-priority alerts into one issue. It escalates to a person only when a decision is needed.
The agent keeps operational memory of service relationships and past results. Engineers can also pull investigations into coding agents over the Model Context Protocol (MCP).
Pricing
Cleric charges in credits at $0.50 each on monthly plans, with unlimited users. Team costs $700 per month (1,400 credits), Pro costs $1,500 (3,000 credits), and Scale costs $3,000 (6,000 credits). Enterprise uses a custom annual credit pool. A free evaluation includes 500 credits with no time limit.
* Disclaimer: Please note that pricing information may not be up to date. For the most accurate and current pricing details, refer to the official website.
Key Features
- ✓
Tracks each change for up to 14 days
- ✓
Links anomalies to the responsible pull request
- ✓
Opens fix pull requests with supporting evidence
- ✓
Deduplicates alerts and closes false positives
- ✓
Stores operational memory of past investigations
Use Cases
Catching Delayed Cache Regressions
A release passes checks on day one, then cache hit rate falls days later. Cleric flags the deviation from baseline. It names the pull request behind it.
Quieting Overnight Pager Noise
Noisy P3 and P4 alerts interrupt on-call engineers. Cleric groups duplicates into one issue and investigates it. Engineers get paged only when a decision is needed.
Diagnosing Payment Service 5xx Errors
A payment service starts returning 5xx errors at night. Cleric rules out deploys, config, and capacity, then traces the cause to one pull request. The escalation arrives with evidence attached.
Reverting Bad Feature Flags Quickly
After a regression, someone must draft a fix. Cleric opens a revert pull request, such as disabling a feature flag. The team reviews the evidence and approves it.
Feeding Coding Agents Production Context
Local coding tools lack production data. Claude Code or Cursor can pull an investigation from Cleric over MCP. The agent then starts from existing findings.
Strengths & Weaknesses
Strengths
Each regression ties to a specific pull request with supporting evidence.
Unlimited users on every plan means team size does not raise cost.
Repeat alerts, retries, and extra passes on the same issue cost no credits.
Access is read-only by default, and Cleric lists SOC 2 Type II compliance.
Customers choose a US or EU hosting region during setup.
Weaknesses
Chats cost one credit per started minute after the first 30 seconds.
Okta and custom SAML or OIDC require the Pro plan or higher.
The Team plan omits scheduled agents and the private network Connector.
Custom data retention and deployment region require an Enterprise contract.
Who Is This For?
Site reliability engineers: Teams that triage alerts by hand can delegate first-pass investigation and keep escalations for real incidents.
Platform engineering teams: Groups supporting many services can verify changes across repositories and apply filters by service.
On-call software engineers: Engineers carrying a pager get fewer low-value pages and arrive at escalations with evidence in hand.
Engineering leaders: Leaders who prefer costs tied to completed work, not seats or tokens, can estimate spend per change and issue.
Frequently Asked Questions
How does Cleric spend credits on its work?
Change verification costs 2 credits per change. Issue investigation costs 10 credits. Chats cost 1 credit per started minute, and requests under 30 seconds are free.
Which changes does Cleric verify by default?
It follows every change announced in Slack release messages. You can add filters to limit verification to certain repositories, services, or changes.
What happens when monthly credits run out?
You can buy more credits at your plan rate or set a hard cap. Cleric does not bill overages without written approval.
Does Cleric train models on customer data?
No. Cleric states it does not use customer data to train Cleric or its underlying models. Its security documentation covers data handling and deletion.
Can Cleric change production without approval?
Access is read-only by default, with write access enabled when you are ready. Cleric proposes fixes as pull requests that a human reviews and approves.
What if an investigation reaches the wrong conclusion?
Corrections and additional passes on the same issue do not use extra credits. A human still decides whether to merge any proposed fix.
How long does setup take?
Cleric says you can connect your stack in an afternoon. Actual time depends on how many code, observability, and chat tools you link.
How do annual Enterprise contracts renew?
Annual contracts run 12 months and renew automatically. The Order Form defines the renewal notice period and true-up terms.
Where does Cleric run its checks and alerts?
Cleric works inside Slack and a web app. It also reads alerts from your connected monitoring stack and posts findings back to your channels.
Cleric integrates with Slack, GitHub, Linear, and Jira, and supports Google Workspace sign-in. Okta and custom SAML or OIDC are available on Pro and above. MCP clients such as Claude Code and Cursor can pull investigations from Cleric.