About Clarm
Clarm builds Clarm Atlas. Clarm Atlas lets non-technical teams build AI coworker agents using building blocks pre-approved by IT and compliance, giving teams the freedom to deploy complete agents straight into production.
What Clarm does
AI coworker agents built in plain language
An operator describes a recurring job, such as reading an order mailbox or checking a claim against policy, and Atlas assembles the agent from approved blocks. Operators never write code, YAML, or prompts full of template syntax, so the person who owns the process is the person who builds it.
Building blocks approved once by IT and compliance
IT and compliance decide which sources an agent may read, which systems it may write to, and which guardrails always apply. Every agent inherits those rules, so the security review happens once per building block instead of once per project.
Agents that run in production on their own
Agents run on a schedule or when an email, message, or file arrives, and they finish the job without a person approving every output. Teams that need a sign-off step for a regulated workflow can add one to that agent only.
Answers with sources and an audit trail
Every answer cites the approved content it came from and every run is recorded. When an auditor asks why an agent did something, the team shows the evidence instead of reconstructing it.
Voice, website chat, and virtual browser agents
The same governed agents can answer on a phone line, in a chat window on your website, or work inside a web application through a virtual browser. Each channel uses the same approved sources and the same audit trail.
What makes Clarm different
Four weeks to a working agent, against a nine-month internal build
A pilot starts with one bounded recurring workflow, your real files, and your current rules. The target is one working agent in production by week four; the usual alternative is a queued engineering project that ships after the process has already changed.
Grants can only narrow access
Authorization is the intersection of what the organisation has approved and what an individual agent has been granted. An agent grant can remove access but can never add access the organisation has not already approved, which is the property IT needs before it lets non-technical teams deploy.
Built for the operator who owns the process
Developer-oriented agent builders expose flow charts, code nodes, or JSON to the person building. Atlas keeps deterministic steps in deterministic blocks and uses a language model only for language, so the operator works in plain words and the output is repeatable.
Deployment from multi-tenant cloud to air-gapped
Atlas runs in the Clarm cloud, in your own VPC, on-premises, or air-gapped, and it can run on Claude, GPT, or a private model on your own hardware. Customer data is never used to train models.
A flat monthly price
The Team plan is $5,000 a month for one workspace and unlimited agents, billed monthly and cancellable at any time. It is not priced per seat or per conversation, so adding a team or a busy month does not change the bill.
Who uses Clarm
- Operations teams at complex, multi-system companies with recurring work still done by hand across email, documents, spreadsheets, and business systems.
- Food, fresh produce, and consumer-goods supply chains: order intake, allocation planning, claims, and shipment recovery.
- Banks and asset managers: client review, document control, and policy updates where every answer must be traceable to a source.
- Healthcare and aviation operations teams with an audit obligation and a backlog of repetitive, rules-based work.
- Heads of Operations, Logistics, Supply Chain, Finance, Quality, and Customer Operations who own a workflow and want to automate it without waiting for an engineering queue.
The team behind Clarm
Clarm started from a pattern inside large companies: the people closest to recurring operational work know exactly how it should be done, and they are almost never the people who can automate it. The usual answer is a six-month engineering project; the other is a general-purpose chatbot that does not know which documents are authoritative or which systems it may touch, so nothing that matters can be trusted to it. Atlas splits the problem: IT and compliance approve the building blocks once, and operators build on top of them.
The company is led by Marcus Storm-Mollard, CEO and co-founder. Marcus studied physics at Imperial College London, built AI systems at Deutsche Bank from 2019 to 2021 and at NielsenIQ from 2022 to 2024, and is a visiting research fellow at the University of Oxford. Find him on LinkedIn and X.
Clarm was founded in 2024 as Dialin AI, Inc., a Delaware corporation, and went through Y Combinator in the X25 batch.
How working with Clarm works
Start with a working session
Bring one recurring process your team runs by hand. We map it onto Atlas blocks live, including which sources and systems it would need approved.
Pilot one bounded workflow
The first commitment is a single workflow with representative files, current rules, and known exceptions. The operator who owns the process builds and runs the agent; IT approves the blocks it uses.
Who you work with
Clarm is a small team, and every message reaches a person rather than a ticket queue. Press and founder questions go straight to the founder, and the response is usually same day.
Channels and support
Sales and pilots through marcus@clarm.com or a booked session; product support through support@clarm.com. We work in English and German.
Self-serve and Enterprise
A Team workspace can be created in minutes with a name and a work email, no card required to start. Enterprise adds multiple workspaces, a self-hosted option, SSO and SAML, a security review, and dedicated support.
Book a working session or read the contact page for every route in.
Clarm key facts
- Company name
- Clarm
- Legal entity
- Dialin AI, Inc. d/b/a Clarm
- Type
- AI agent platform for non-technical operations teams
- Core offering
- Clarm Atlas
- Founded
- 2024
- Backed by
- Y Combinator (X25)
- Registered office
- Dover, Delaware, United States
- Registered address
- Dialin AI, Inc., 1111B S Governors Ave STE 29410, Dover, DE, 19904, US
- Website
- https://clarm.com
- Pricing
- Team plan $5,000/month (one workspace, unlimited agents); Enterprise custom
- Contract terms
- Team plan billed monthly, cancel anytime; Enterprise by agreement
- Deployment
- Clarm cloud, customer VPC, on-premises, or air-gapped
- Communication
- Email (marcus@clarm.com, support@clarm.com) and booked working sessions; English and German
- Industries served
- Food and consumer-goods supply chains, banking and asset management, healthcare, aviation operations
- Alternatives compared
- Microsoft Copilot Studio, Glean, Sierra, Lindy, and other agent platforms
- Social
- Y Combinator 路 LinkedIn 路 X 路 GitHub
Frequently asked questions about Clarm
What is Clarm Atlas?
Clarm Atlas lets non-technical teams build AI coworker agents using building blocks pre-approved by IT and compliance, giving teams the freedom to deploy complete agents straight into production.
Does a person have to approve everything a Clarm agent does?
No. Agents run autonomously inside the rails IT and compliance approved upstream. Teams in regulated workflows can add an approval step to a specific agent when their process requires one.
Do I need to write code to build an agent in Clarm Atlas?
No. Operators describe the job in plain language and build on a visual canvas. They never see code, YAML, or JSON.
Does Clarm train AI models on customer data?
No. Customer data is never used to train models. Deployment options run from the Clarm cloud through your own VPC to on-premises and air-gapped.
How much does Clarm Atlas cost?
The Team plan is $5,000 per month for one workspace and unlimited agents, billed monthly and cancellable at any time. Enterprise pricing is custom and adds multiple workspaces, self-hosting, SSO, and a security review.
Where can I find Clarm security and compliance documentation?
Current assurance artifacts, subprocessors, and deployment details are published on the Clarm Trust Center at compliance.clarm.com. Anything it does not cover is answered during procurement.
Where to go next
Security and compliance evidence is on the Trust Center. To integrate with Clarm, start at the developer hub. For how we handle data, read the privacy policy and the terms of service.