
Serval · Serval: AI Help Desk Agent and Workflow Builder for IT, HR and Internal Service Management
Serval is an AI-native IT service management platform with two separate agents: Catalyst builds code-based workflows from plain-English prompts, and the Help Desk Agent resolves employee IT, HR, security and finance requests in Slack, Microsoft Teams, email or a portal by running only workflows admins have approved. It is aimed at IT and people-operations teams at mid-size and large companies that want to automate password resets, access requests, onboarding and other internal tickets, either replacing ServiceNow or Jira Service Management or working on top of them.
Best for
IT, people-operations and security teams at mid-size and large companies that handle a steady volume of repetitive internal requests in Slack or Teams and want an AI agent to resolve them through approved, auditable workflows, whether replacing ServiceNow or Jira Service Management or layering on top of it
Not ideal for
Teams looking for a customer-facing support agent, small companies that want a self-serve tool with a published price, and organizations unwilling to invest time in reviewing and maintaining their own workflow library
Who it's for
IT service desk leads, IT and security engineers, people-operations teams and internal support functions (finance, legal) at mid-size and large companies
Serval's most useful design choice is the split between an agent that builds and an agent that acts. The employee-facing agent can only call workflows a person has reviewed and published, and those run as versioned code with run logs, which gives security and audit teams a clear answer to what the AI is allowed to do. The catch is that coverage is only as good as the workflow library, so the real question in a pilot is how much of your ticket volume Catalyst can turn into approved workflows, and how much review work that creates for your admins. Serval is a serious option if you are weighing a move off ServiceNow or Jira Service Management, but you do not have to decide that up front: the automation-layer mode lets you test it on your noisiest request types first. Expect a sales-led, pilot-first rollout with no public price.
Who should use it
IT and people-operations leaders handling high volumes of repetitive internal requests (access, passwords, devices, onboarding, PTO) who want them resolved in Slack or Teams through auditable workflows, and teams reconsidering ServiceNow or Jira Service Management licensing.
Who should skip it
Companies looking for a customer-facing support agent, small teams that need a self-serve product with a published price, and organizations that need fully self-hosted deployment outside AWS.
Full-stack ESM
Contact sales
Automation Layer
Contact sales
Note: Serval publishes no price list. Its pricing page says every engagement starts with a pilot (plan, pilot, deploy, go-live) and that the proposal is built from it.
MFA reset in Microsoft Teams
An employee with a new phone asks in Teams; the help desk agent confirms identity through the authenticated session, runs the published Entra MFA reset workflow and logs the run on a ticket without an IT engineer stepping in.
Just-in-time app access
The agent asks why the employee needs access, matches the reason to a role, routes it for approval and grants access that expires automatically, with the decision recorded for later access reviews.
Migrating automations off ServiceNow
Catalyst reads exported ServiceNow tickets and SOP documents, ranks the most common manual requests and drafts workflows for them, which admins test and publish while Serval syncs with ServiceNow until cutover.
Serval vs. Rovo
Rovo is the AI layer built into Atlassian Cloud, including service agents for Jira Service Management, metered in Rovo credits included with Atlassian plans. It suits teams that want to stay on JSM. Serval is a separate ITSM that can replace JSM or sync with it, and it adds built-in just-in-time access, access reviews and license reclamation, with Catalyst writing workflows as code rather than Jira automation rules.
Serval vs. Agentforce
Agentforce includes Paige, a Salesforce agent for employee IT and HR requests, built on Salesforce data, Flows and Agentforce pricing. It fits companies whose processes already run on Salesforce. Serval is a dedicated service management platform whose help desk agent acts only through published TypeScript workflows, and it can replace or sit on top of ServiceNow or Jira Service Management.
Serval vs. Nexus EDGE
Nexus EDGE from Yellow.ai pairs a cloud agent with a desktop client that runs fixes on the employee's own machine after the employee approves each change. Serval works from the service desk side instead: it runs approved workflows against connected systems such as the identity provider, HRIS and MDM, and includes ticketing, ITSM processes and access management.
What is Serval?
Serval is an AI-native IT service management (ITSM) platform for internal employee support. It combines two AI agents with a full service desk: Catalyst, which builds automations from natural-language prompts, and the Help Desk Agent, which resolves employee requests in Slack, Microsoft Teams, email or a web portal. It also includes ticketing, change and incident management, a knowledge base, asset management, a CMDB and access management, and is used by IT, HR, security, finance and legal teams.
What is the difference between Serval's Catalyst and the Help Desk Agent?
Catalyst is for admins. It writes workflows as TypeScript, plus skills, journeys, access policies and dashboards, from plain-English descriptions, documents or ticket history, and stages them for review before publishing. The Help Desk Agent is for employees. It answers questions from connected knowledge and takes action only by running workflows that admins have published. Serval's documentation says the two agents do not share context and the help desk agent cannot create or edit workflows.
How much does Serval cost?
Serval does not publish prices. Its pricing page says every engagement starts with a pilot to prove value, after which Serval builds a proposal that matches how your team operates. You can buy it as a full-stack enterprise service management platform or as an automation layer over your existing one. The pricing page lists deployment engineering support, post-deployment success help and training among what is included, but no seat, usage or contract terms are public.
Does Serval replace ServiceNow or Jira Service Management?
It can do either. Serval positions itself as a full replacement for ServiceNow, Jira Service Management and Freshservice, with request, incident, problem and change management, a CMDB and asset management built in. It can also sit on top of an existing ITSM as an automation layer with two-way sync, so teams can automate requests first and move the system of record later.
Can Serval be self-hosted?
Yes, in two ways. Hybrid Self-Hosted runs only the Serval worker on a Linux host in your network, so workflows can reach private systems such as on-premises Active Directory, while Serval runs the backend in its cloud. BYOC on AWS runs the full application single-tenant in an AWS account you own, operated by your team or by Serval, and requires your own Anthropic and OpenAI API keys. Step-by-step deployment instructions are shared with customers during onboarding. The default is the cloud service, with US or EU data residency.
Is Serval secure, and can the AI take actions on its own?
Serval's documentation says the help desk agent can act only through published, deterministic workflows, that permissions and approvals are enforced in code rather than by the model, and that integration credentials are injected server-side so workflow code and the model never see them. Background agents need explicit admin approval for write actions unless an admin pre-approves specific workflows or turns on a bypass mode for that agent. Serval's documentation states it is SOC 2 Type II certified, with TLS 1.3 in transit, AES-256 at rest and regular penetration testing.
Can I use Serval from Claude or other AI assistants?
Yes. Serval runs an MCP server per region that Claude (claude.ai and Claude Desktop) and other MCP clients can connect to after approving access in the browser. Through it, employees can create tickets, run workflows, check request status and look up people. Your admin may need to enable MCP access first.
Atlassian
Organizations already running projects, documentation and service work in Jira, Confluence and Jira Service Management cloud that want AI search, chat and agents grounded in that work, built and governed by Atlassian admins without a separate AI rollout
PaidSalesforce
Companies whose customer data, service processes, and sales workflows already run on Salesforce and that want agents acting directly on that data across service, sales, commerce, and employee support
Paid
Yellow.ai
Large enterprises with a heavy L1 IT and HR ticket load that want an employee-facing agent able to fix common device issues on the machine itself, within each employee's existing permissions and with a single audit trail
CustomServal splits the work that AI help desks usually blend together. Catalyst is the agent for admins: you describe an automation in natural language, or hand it ServiceNow ticket history, an SOP document or an onboarding spreadsheet, and it writes the workflow as TypeScript, along with skills (instructions for how the help desk agent handles a type of request), multi-step journeys such as onboarding and offboarding, access-request policies and dashboards. Nothing goes live until someone reviews, tests and publishes it, and admins can require approval before publishing. The Help Desk Agent is the one employees talk to. It answers questions from connected knowledge (Confluence, SharePoint, Google Drive, Notion, ServiceNow or Serval's own knowledge base) with citations, asks follow-up questions, and takes action only by invoking published workflows: resetting a password or Microsoft Entra MFA, granting time-limited app access, booking PTO in the HRIS or fixing a device through MDM. Serval's security documentation says the two agents do not share context, the help desk agent cannot create or edit workflows, and a published workflow runs as fixed code with no LLM in the path unless your team adds one, with every run logged step by step. Background agents run on a schedule to find gaps, for example manually resolved tickets that could be automated, failing workflows, Macs without escrowed FileVault keys or unused licenses, and draft fixes for an admin to approve. Around the agents sits a full ITSM: ticketing with SLAs and routing, request, incident, problem and change management, a knowledge base, a service catalog, hardware and software asset management, a CMDB, access reviews and license reclamation. Serval can also run as an automation layer over an existing ITSM with two-way sync, so teams can start without migrating. The tradeoffs: pricing is not published and every engagement starts with a pilot and a deployment team, the product is built for internal employee support rather than customer-facing service, and its value depends on building and maintaining a library of approved workflows. Serval raised a $150 million Series C led by Sequoia at a $3 billion valuation on October 29, 2026.
Serval announced a $150 million Series C at a $3 billion valuation, led by Sequoia, its third round in 12 months. The company positions the funding around replacing ServiceNow as the system of record and launched ProServal, a professional services arm of automation engineers who work inside customer teams.
Serval added ad-hoc approvals on any ticket, a redesigned change ticket experience, email from a company's own domain, pre-filled forms and user-specific database rules.
Catalyst, Serval's automation agent, became generally available and on by default for every team, replacing the standalone workflow builder. It drafts workflows, skills, dashboards, journeys and access setups from one chat, including automations suggested from ticket history.
Are you the founder? Claim this listing →