Built by Practitioners. Designed for every
On-Call Engineer.
Sherlocks AI was born from lived experience transforming painful incident nights into efficient, AI-powered investigations.
Sherlocks AI is an AI SRE platform that investigates production incidents end to end, across the observability stack you already run, for engineering teams who carry a pager.
What Sherlocks AI does
Incident investigation and root cause analysis
When an alert fires, Sherlocks AI reads your metrics, logs, traces, deploy history and cluster state together, tests competing explanations, and posts the cause with the evidence behind it. You get a root cause with the working shown, not a summary of the alert you already had. Read more.
Postmortem drafting
The timeline is assembled from system data rather than memory, including the hypotheses that were tested and ruled out. The draft is ready by the time the incident closes, so the review meeting starts from evidence instead of recollection.
Daily reliability reviews
Scheduled checks written in plain English run against your systems every morning and report what changed. The point is to catch the slow problems, the ones that look normal day to day, before they become an incident. Read more.
Runbooks and remediation steps
Every investigation ends with the exact commands to run, for a person to review and run. Sherlocks AI is read-only by default and does not execute changes in your production systems.
Onboarding and knowledge retention
The knowledge graph keeps a map of your services, dependencies and past incidents, and corrections made in Slack are folded back into it. New on-call engineers inherit that context instead of rebuilding it. Read more.
What makes Sherlocks AI different
The data agent is read-only and runs inside your VPC
Watson, the component that touches your systems, has zero write, execute or delete permissions, and it runs inside your cloud rather than ours. Any change to production needs explicit human approval through a channel the agent cannot reach.
It works with your observability stack instead of replacing it
Sherlocks AI plugs into Datadog, New Relic, Prometheus, Kubernetes and cloud APIs. Tools like Datadog Bits investigate only within their own platform, so switching vendors is the price of entry. Here it is not, and nothing needs re-instrumenting.
Multiple LLM providers, not one
Sherlocks AI runs on Anthropic Claude, OpenAI, AWS Bedrock, or open-weight models for teams that cannot send data to a hosted provider. Vendors built Claude-first inherit their provider's outages; here a bad afternoon at one provider does not stop an investigation.
A free tier you can actually run an incident on
30 investigations a month at no cost, no credit card, self-serve. Most of the category, including Resolve AI and Traversal, has no public pricing at all and starts with a sales cycle.
The investigation record is auditable
Every hypothesis tested, every source read and every explanation ruled out is written down, and where the answer needed a system Sherlocks AI could not reach, it says so instead of guessing. Twenty full investigations are published at sherlocks.ai/examples so you can judge the quality before you buy.
For the full market view, including tools we did not build, see our comparison of AI SRE tools and the eight questions to ask any vendor before buying.
Who uses Sherlocks AI
- Platform and SRE teams running production on Kubernetes, AWS, GCP or Azure, with an on-call rotation already in place.
- Engineering teams at 50 to 500-person product companies where one squad owns reliability for many services.
- Data and infrastructure teams whose incidents cross the pipeline, the database and the cluster in a single outage.
- Teams already paying for Datadog, New Relic, Prometheus or Grafana, where detection is fast and investigation is the slow part.
- Regulated teams in finance, healthcare and the public sector that need the agent to run inside their own network.
- Managed service providers and support organisations handling L1 and L2 for customer estates.
The team behind Sherlocks AI
Gaurav Toshniwal, co-founder and CEO
Gaurav leads product and go to market at Sherlocks AI and writes most of the engineering guides published here. LinkedIn.
Akshat Sandhaliya, co-founder and CTO
Akshat leads engineering, including the investigation agents and the read-only Watson data agent that runs inside customer environments. LinkedIn.
How the company started
Sherlocks AI is a team of 8. It was founded in 2025 after one outage made the problem concrete: detection was fast, and the hours went into working out why. The team has since raised a $900K pre-seed round and built the product around that gap, the stretch between the alert firing and somebody understanding the cause.
The People Behind Sherlocks AI
Our Journey So Far
The Incident That Started It All
A major outage made us realize that incidents shouldn't drain people. We imagined a smarter, faster way to resolve problems automatically, saving time and reducing stress.
How working with Sherlocks AI works
Where it runs
The Watson data agent is deployed inside your own VPC with read-only credentials, or the whole product can be self-hosted. Teams that cannot send data to a hosted model provider can run it against open-weight models.
Where you talk to it
Slack. Investigations appear in the channel the alert arrived in, and corrections you make in the thread are folded back into the knowledge graph for the next incident. There is a web app for the full investigation record.
Getting started
The free plan is self-serve: connect your observability tools and your cluster, and the first investigations run on the next alert. No code changes and no re-instrumentation are required. Enterprise deployments are set up with the team directly.
What to expect from an investigation
An investigation runs the moment the alert fires, and the root cause arrives in minutes rather than the hours a manual correlation takes. It comes with the evidence behind it, the hypotheses that were ruled out, and the commands to fix it.
Who you deal with
It is a small company, so support comes from the people who build the product, and the founders are reachable directly. Contact us or email gaurav@sherlocks.ai.
Key facts about Sherlocks AI
| Company name | Sherlocks AI |
|---|---|
| Type | AI SRE platform for production incident investigation. Private company, pre-seed funded. |
| Founded | 2025 |
| Founders | Gaurav Toshniwal (CEO) and Akshat Sandhaliya (CTO) |
| Headquarters | 8 The Green, Ste A, Dover, Delaware 19901, United States |
| Website | https://www.sherlocks.ai |
| Core offering | AI-led incident investigation across your existing observability stack, using 16+ domain-specialized agents and a read-only data agent that runs inside your VPC. |
| Pricing | Free: $0 for 30 investigations per month. Pro: $500 per month. Enterprise: custom. See pricing. |
| Contract terms | Free and Pro are monthly and can be cancelled at any time. Enterprise terms are agreed per contract. |
| Services | Incident investigation and root cause analysis, postmortem drafting, daily reliability reviews, runbooks and remediation steps, onboarding and knowledge retention. |
| Integrations | Datadog, New Relic, Prometheus, Grafana, Kubernetes, AWS, GCP, Azure, PostgreSQL, MySQL, MongoDB, Redis, Kafka, RabbitMQ, GitHub, Jenkins, Slack. |
| Communication | Slack-native. Investigations arrive in the channel the alert arrived in. |
| Time to root cause | Minutes from the alert, rather than the hours manual correlation takes. |
| Team | A team of 8. |
| Funding | $900K pre-seed (₹7.5 Cr) |
| Security and compliance | SOC 2 Type 2. Read-only access. PII stripped before anything leaves your environment. |
| Notable customers | Fynd, Topmate, Wafeq, TradeIndia, Lokal |
| Competitors | Resolve AI, incident.io, Traversal, Datadog Bits Investigation, NudgeBee, Rootly AI |
| Rating | 4.9 out of 5 from 28 reviews |
| Social | LinkedInXYouTubeG2Crunchbase |
Frequently asked questions
What is Sherlocks AI?
Sherlocks AI is an AI SRE platform that investigates production incidents across the observability stack a team already runs. When an alert fires it reads metrics, logs, traces, deploy history and cluster state together, tests competing explanations, and posts the root cause with the evidence behind it.
Who is Sherlocks AI for?
Engineering and SRE teams running production systems on Kubernetes, AWS, GCP or Azure, typically with an on-call rotation and an observability stack already in place. It suits teams where investigation, not detection, is the slow part of an incident.
Does Sherlocks AI replace my observability tools?
No. It sits on top of Datadog, New Relic, Prometheus and the rest, and reads what they already collect. Nothing needs re-instrumenting and no telemetry has to move.
Can Sherlocks AI make changes to my production systems?
No, not on its own. The Watson data agent is read-only and has zero write, execute or delete permissions. It produces the exact commands to run, and a human approves and runs them.
Where does my data go?
The Watson data agent runs inside your own VPC, and sensitive patterns are stripped before anything leaves your environment. Teams that cannot send data to a hosted model provider can run Sherlocks AI with open-weight models instead.
How much does Sherlocks AI cost?
The free plan covers 30 investigations a month with no credit card. Pro is $500 per month, and Enterprise is priced per contract. Free and Pro are monthly and can be cancelled at any time.
Who founded Sherlocks AI?
Gaurav Toshniwal, the CEO, and Akshat Sandhaliya, the CTO. The company was founded in 2025 and has raised a $900K pre-seed round.
How is Sherlocks AI different from Resolve AI, incident.io or Traversal?
The main differences are deployment and transparency: the data agent is read-only and runs inside your own VPC, it works across several LLM providers rather than one, the free tier is self-serve with public pricing, and twenty complete investigations are published so the quality of the reasoning can be checked before buying.
Trusted By Leading Engineering Teams
Mission
Empowering engineering and security teams to move beyond the manual grind with intelligent tools that make incident response effortless, faster, and more human focused.
Vision
Transforming the global incident management landscape from a stressful, manual burden into a seamless, autonomous experience that allows engineers to focus on creation.
What We Stand For
Customer obsession
We place customer at the heart of everything we do responding to every tickets in 24 hrs to make sure no quivery goes unanswered.
Bias For Action
We value speed without compromising quality, prototyping new ideas within a week to gather feedback and iterate faster than before.
Transparency
Openness fosters trust, so all internal dashboards and reports are accessible company-wide to everyone and helping, empowering informed decisions.
Trust In AI
We leverage AI responsibly to augment human intelligence, providing recommendations that are always reviewed by experts before action.
What Engineering Leaders Say
Sachin Mittal
CEO
"As CEO of InterviewVector, I need to trust that our platform is reliable. Sherlocks AI empowers our entire team, not just senior engineers, to understand and resolve issues quickly. The AI-driven insights mean we catch problems before they impact our users, and the Slack integration keeps everyone in the loop without endless meetings."
Harsh Sarohi
Head of Engineering
"At TradeIndia, our systems are vast and complex, with multiple teams and responsibilities. Sherlocks AI brings consistency for incident response across departments. Its AI-driven insights not only detect potential issues early but also provide actionable steps, saving us valuable time. The direct Slack integration is excellent as our engineers can get answers to stack-specific and historic queries instantly."
Dinesh Singh
Co-founder & CTO
"Sherlocks integrates seamlessly with Topmate’s tech stack. It’s excellent at clarifying whether incidents stem from our backend infrastructure or application layer, so tickets aren’t bouncing between teams and support resolves issues much faster. Triggering investigations directly from Slack via @sherlocks is a real time-saver."
Mayank Shekhar
Co-Founder & CTO
"As CTO at Robylon, our stack is highly distributed and event-driven. Sherlocks integrates easy, auto-classifies faults across services, and drastically reduces the volume of “noisy” alerts that our pipelines used to emit. Its RCA engine surfaces the actual sources of failures, so our MTTR is down, and reactive firefighting has become a rare event."
Shubhanshu Prasi
Senior Product Manager
"At Repos Energy, customer trust is everything. Any downtime directly impacts our users' experience and our reputation. Sherlocks AI has been instrumental in helping us stay ahead of issues before they affect customers. The platform's proactive monitoring and intelligent alerting mean we're catching problems early before they escalate into real outages."
Gaurav Verma
CEO
"The platform’s integrations were seamless and had us running almost instantly. Sherlocks’s AI can proactively detect outages and identify the root cause. We especially appreciate how follow-up investigations can be triggered directly from Slack."