OpenHands runs coding agents in engineering; Kortix is open source and governs the same work after it leaves engineering. Who owns the rules, the approvals and the record when an agent's output reaches other teams? Kortix is the open-source AI Operating System: your agents, their skills, your company memory, and every connector in one git repo you own.
Each layer here is a real Kortix excerpt or command from Kortix's docs.
One repo is the company: agents, skills, memory, connectors and triggers are files, versioned and reviewed like code.
kortix_version: 2default_agent: kortixagents: kortix: file: agents/kortix.md connectors: all3,000+ apps in a click, plus MCP, OpenAPI, Postman, GraphQL and raw HTTP.
policies: - match: "get_*" action: always_run - match: "*" action: require_approvalApproval gates you set. Off until you set them.
policy: default_mode: risk policies: - match: gmail.send_email action: require_approvalPer-resource permissions for people and agents. Roles, groups, and an audit trail.
$ kortix members invite alice@example.com --role member$ kortix access grant --user alice@example.com --agent kortix$ kortix access grant --everyone --agent kortixEvery action is recorded. Reading, exporting and streaming the audit log depend on the plan.
Every account action and every agent action is captured on every plan.The plan decides who may read, export, or stream that record.Run it on Kortix Cloud, in your VPC, or on your own on-prem network. Self-host is free.
$ kortix self-host init --domain kortix.example.com$ kortix self-host startOpenHands is a self-hosted developer control center for coding agents and automations. Its Agent Canvas connects to one or more agent backends and can run the OpenHands agent or a third-party agent such as Claude Code, Codex, Gemini CLI or OpenCode across local, remote and cloud environments (OpenHands on GitHub).
OpenHands Cloud is the managed service for a team that would rather not operate its own backend and sandbox infrastructure, and OpenHands Enterprise is the commercial tier for organizations that need their agents to run inside their own network (OpenHands docs).
Kortix is open source, and it carries the same class of agent work in a repo the whole company owns. OpenHands organizes engineering work from issue to pull request. Kortix is built for the point where an agent's output reaches teams outside engineering, because that is when a rule per tool call, an owner for every approval, an identity for every agent and one durable record stop being optional. The OpenHands vs Kortix page holds the head-to-head between the platforms.
A connector is one governed path from a project to one external app, MCP server or API. Kortix stores the credential, picks the account, applies the project's policy and writes an audit event for every call. Connector credentials are brokered server-side and never enter the machine (Read the docs).
Every connector action resolves to one of three policies: always_run, require_approval or block, shown in the product as Allow, Ask and Block. A rule matches a glob over fully-qualified tool paths, and a condition can match the arguments, so a rule can mean "send mail only to these addresses". An Ask holds the call with its arguments in front of a person, and the agent resumes after approval.
Approval gates you set, and they are off until you set them. The manifest's policy.default_mode falls back to allow_all, so the rules are a deliberate block in kortix.yaml, versioned with everything else (Kortix docs). OpenHands Enterprise describes prompt-injection detection with an automatic halt and a hold on risky actions for human review (OpenHands Enterprise).
Kortix treats people and agents as principals. Each person holds one account role, plus a project role on each project they work in, and a member needs a grant for at least one agent before the API lets them start a session. Per-resource permissions for people and agents. Roles, groups, and an audit trail (Kortix docs).
Kortix covers enterprise identity with SAML 2.0 single sign-on and SCIM 2.0. A self-hosted instance adds role-based access control and audit in its enterprise tier. OpenHands Enterprise lists SAML and SSO support and multi-user role-based access control (OpenHands pricing, checked October 2026).
Every action is recorded. Reading, exporting and streaming the audit log depend on the plan. Kortix keeps 90 days of events live and an archive up to 365 days (Kortix security).
OpenHands Enterprise says every run is logged, attributed and governed by a team's cost and security policies (OpenHands Enterprise). A team comparing the two on audit still has to answer one question: where the log lives and who can read it.
Run it on Kortix Cloud, in your VPC, or on your own on-prem network. Self-host is free and it is one Docker Compose stack that the kortix self-host commands install, configure and update. Sessions run on a cloud sandbox provider, so the instance must be reachable from the internet for the sandbox to call back; this is not an offline deployment (Kortix docs).
OpenHands runs in a team's own environment, on-premises or in a private cloud, and OpenHands Enterprise is self-hosted in a team's VPC on Kubernetes (OpenHands, OpenHands Enterprise).
Configuration in Kortix is a git repo the company owns. Agents, skills, memory, connectors, triggers and policy are files, so a team clones the repo, diffs any change and keeps the whole system even if it stops using the managed cloud. If the work stays inside engineering, OpenHands fits that lane as a self-hosted control center for coding agents. When the same agents start to touch support, finance and sales, the rules and the record need one home across all of them, which is what Kortix is built for. See the OpenHands alternatives and Kortix pricing.
Yes. OpenHands stays the coding agent for engineering work, and Kortix runs the same class of agent work across the company's tools with per-tool rules and approvals. Kortix does not replace the coding agent. It owns the rules, the identities and the record once work leaves engineering.
In Kortix, agents, skills, memory, connectors, triggers and policy are files in one git repo the company owns. A team clones the repo with kortix projects clone, reviews every change as a diff, and keeps the system even if it stops using the managed cloud.
Free is $0 with 200 credits per month and 1 project. The team plan is $40 per seat per month with 2,500 credits per month per seat, pooled, and self-host is free. Prices checked October 2026 (Kortix pricing).
Set the rules, keep the approvals, and hold the record in a repo you control.