Companies

Decawork

decawork.ai

Decawork provides IT teams a governed control plane for deploying, monitoring, and maintaining enterprise AI agents.

HQSan Francisco, California, United States
Employees1-50
Jobs checked 16h ago
CybersecurityAI InfrastructureB2B SaaS

About

Decawork builds an agent control plane for IT teams to deploy, govern, monitor, and maintain AI agents across their companies. Its differentiation is governance-focused deployment: agent-specific identities and scoped credentials, IT approval before launch, continuous monitoring, access controls, and auditable actions.

Market

Decawork competes in the emerging B2B enterprise AI-agent infrastructure and governance market, addressing the operational risk created when many teams build agents that touch real systems and data. It positions itself as an IT-owned control plane spanning deployment, identity, scoped credentials, approval-gated tool calls, monitoring, maintenance, and audit, rather than only an agent builder or observability product. Its differentiation is cross-stack, lifecycle-wide governance for the agent fleet, while Zenity emphasizes agent security and governance, LangSmith emphasizes managed deployment and observability, and cloud platforms provide vendor-specific managed agent services.

Target Customers

Decawork targets B2B organizations with multiple teams building and deploying AI agents, particularly companies where centralized IT inherits responsibility for the agent fleet and its access to business systems. Its primary buyers and users are IT administrators, IT operations, and security or governance owners; the company also describes its platform as supporting governed AI co-workers for non-engineering teams.

At a Glance

Problem

As companies move from isolated AI experiments to many agents used by different teams, IT inherits an unmanaged fleet. Agents increasingly touch real systems and sensitive data, but deployment, access control, monitoring, and maintenance are fragmented. This leaves IT choosing between ignoring the risk—allowing security, compliance, and reliability problems to accumulate—or blocking agents and eliminating the productivity gains they promise.

The clearest use case is an IT team governing departmental agents that perform recurring work across enterprise systems, such as finance invoice reminders or vendor onboarding. The economic pain is the widening gap between the low cost of creating an agent and the much higher operational cost of safely running, troubleshooting, approving, and auditing dozens or hundreds of them.

Product / Service

Decawork positions itself as an agent control plane: one place for IT teams to deploy, govern, and maintain the AI agents operating across the company. Each agent receives its own identity and scoped credentials rather than borrowing a person’s access; IT reviews and approves it before go-live, while the platform routes tool calls through a gateway, gates sensitive actions, and keeps an audit trail of what happened.

After deployment, Decawork monitors every run, alerts IT when credentials expire or tools change, provides a recovery path, and can retire unused agents. The delivery model appears to be an enterprise, founder-led product sold through a demo rather than a self-serve offering; the benefit is a standardized operating layer that lets IT enable agent productivity without surrendering visibility, authority, or control.

Market

Decawork competes in the emerging B2B AI-agent infrastructure and governance market, specifically the control-plane layer for enterprise agents. The adjacent competitive set includes Airia, which is also positioned as a model-agnostic orchestration and governance control plane, as well as broader AI-governance platforms such as OneTrust, Arthur, Credo AI, Fiddler, and Microsoft Foundry. Decawork’s differentiation is its narrow operational focus on agent identity, access, approvals, runtime monitoring, and auditability for IT teams rather than general model-risk or AI-compliance management.

The company appears very early. Y Combinator lists it as founded in 2026, active in its Summer 2026 batch, with a two-person team and no listed jobs; LinkedIn describes a two-to-three-person company. Its public site emphasizes booking a founder demo and does not disclose customers, revenue, or deployment metrics, so the best-supported characterization is an early commercial or likely pre-revenue startup rather than a company with publicly demonstrated traction.

Founders & Leadership

Sarthak AggarwalFounder
Co-Founder
Aman RajFounder
Co-Founder

Funding History

2026-07
Seed$500K

Y Combinator

Recent News

2026-07-28
Decawork: The Agent Control Plane for IT teams

Y Combinator’s company profile presents Decawork as an active Summer 2026 AI and B2B infrastructure startup. It provides IT teams with one place to deploy, govern, and maintain the AI agents operating across their company.

2026-07-15
Decawork — Y Combinator Summer 2026 AI directory listing

Y Combinator’s AI startup directory listed Decawork as a Summer 2026, active startup with two employees in San Francisco. The listing describes its product as a centralized control plane for deploying, governing, and maintaining enterprise AI agents.

Active Roles

0

No active roles right now.

Get notified when they post

Business Model

Decawork appears to monetize as a demo-led B2B enterprise-software platform, selling IT teams access to its AI-agent control plane. The company publicly promotes founder-led demos but does not disclose pricing or whether contracts are subscription-, usage-, or agent-based.

Products

Agent control plane for enterprise AI agentsAgent deployment and go-live management, including identity provisioning and scoped accessRuntime maintenance, monitoring, alerting, and agent retirementGovernance and security gateway with approval controls for sensitive actionsAgent-fleet inventory, audit trails, and operational oversightAI agents for IT service requests, access management, and natural-language automations

Customers

No publicly named enterprise customers identified

Tech Stack

AI-agent runtime and agentic automationPer-agent identities with scoped credentialsPolicy gateway for tool calls and approval gatingLive monitoring, alerting, run-level auditing, and agent-fleet controlsEnterprise integrations with Okta, Google Cloud, GitHub, Notion, Slack, and Jira

Competitors

Zenity
LangSmith
Microsoft Foundry Agent Service
Amazon Bedrock AgentCore