Kortix
Open-source AI Management System that puts your agents, skills, company memory, and connectors in a git repo you own, delivering real work on cloud computers.
- Source repo
- kortix-ai/suna
- Stars
- ★ 20k
- Last updated
- today
- License
- NOASSERTION
- Primary language
- TypeScript
- FA score
- 60/100 · Some gaps
At a glance
- How it runs
- Works with
- Universal · cross-platform
- Cost
- Free tier plus a paid hosted plan
- Setup effort
- Medium · a few setup steps
- You'll need
- Typical use
- A development team leader wants to version-control their AI workflows and own an auditable repository defining agents and processes.
- Not a fit if
- Teams that need a fully offline or air-gapped install
- Teams wanting turnkey setup without CLI and git workflows
- Users needing Teams, email, or voice as production channels
- Source review
- 60/100 · Some gaps
What does this agent do, and when should you use it?
Kortix is an open-source AI Management System positioned as an alternative to Claude Cowork and ChatGPT Work. It centralizes a company's agents, skills, memory, and connectors in a single versioned git repository owned by the company. Each session launches an isolated cloud computer (sandbox) on its own branch, where agents work and propose changes via a change request that a human approves before merging to main. The system supports any model provider and your own API keys, and can be deployed on Kortix Cloud, your own VPC, or on-premises. It includes a CLI for managing sessions, change requests, and chat, and integrates with Slack for team collaboration.
Kortix runs OpenCode agents that operate on real cloud computers—isolated sandboxes per session, each on a branch named after the session. Agents can install software, run commands, and access tools, then submit work through change requests for user review and merge. The platform manages agents, skills (reusable know-how), memory (plain files), connectors (3,000+ apps via MCP, OpenAPI, GraphQL, HTTP), secrets (server-side brokering, encrypted at rest), and triggers (cron, signed webhooks). It provides CLI commands like kortix sessions new, kortix cr ls, and kortix chat, and supports Slack integration. Deployment uses Docker and can be initiated with kortix self-host start.
- A development team leader wants to version-control their AI workflows and own an auditable repository defining agents and processes.
- A solo developer self-hosts AI workflows entirely on their laptop or VPS, using their own API keys.
- A company needs to run thousands of parallel, isolated agents across many sessions, for tasks like research or code generation.
- An operations team wants to trigger sessions from Slack, starting tasks and receiving results without leaving their chat app.
- A security-sensitive organization requires strict permissions, audit trails, and deny-by-default merging of agent changes.
How do you install or deploy this agent?
Clone the repository: git clone https://github.com/kortix-ai/suna.git. Requires Node.js and pnpm. Run pnpm install to install dependencies. For local development, use pnpm dev to start the web app and API. For self-hosting, run kortix self-host start, which pulls Docker images and walks through setup, asking for integration credentials. The CLI can be installed via curl -fsSL https://kortix.com/install | bash.
How do you use this agent?
- Initialize a project with
kortix init, generatingkortix.yamlplus agent, skill, and runtime configuration. 2. Deploy withkortix ship. 3. Start a session withkortix sessions new --prompt "...". 4. Review change requests withkortix cr lsand merge them. 5. Chat with the agent usingkortix chat. Alternatively, sign up for Kortix Cloud and create a project with no installation.
What are this agent's strengths and limitations?
- Fully open-source and self-hostable, with data, models, and infrastructure under your control.
- Each session runs on an isolated cloud computer, enabling parallel execution of thousands of sandboxes.
- Company knowledge, configuration, and changes are versioned, diffable, and auditable.
- Model-agnostic, supporting any provider and your own API keys.
- Large integration ecosystem: 3,000+ apps, plus MCP, OpenAPI, GraphQL, and HTTP connectors.
- Self-hosted setup requires Docker and Docker Hub access; not fully offline.
- Initial setup requires familiarity with CLI and git workflows compared to turnkey products.
- Memory system is file-based; compound memory is planned but not yet implemented.
- Slack is the only production-grade channel; Teams, email, and voice are experimental or planned.
How does this agent compare with similar options?
Compared to Claude Cowork and ChatGPT Work, Kortix is open-source, supports any model provider and your own API keys, and can be deployed on your own infrastructure. Claude Cowork is limited to Anthropic models and their cloud, while ChatGPT Work is limited to its own models and cannot be self-hosted.
Key facts side by side with the most closely related agents.
| Agent | Source review | Form / cost | Stars | Updated | Language | Full support on |
|---|---|---|---|---|---|---|
| Kortix This agent | 60 · Some gaps | CLIFreemium | ★ 20k | today | TypeScript | — |
| SandBase Harness | 52 · Major gaps | CLIFree + model costs | ★ 639 | 7d ago | TypeScript | OpenAI API · Claude API |
| OpenOps | 59 · Major gaps | Self-hosted serviceFreemium | ★ 1.1k | today | TypeScript | — |
| Agent Barn | 55 · Major gaps | Self-hosted serviceFree + model costs | ★ 110 | 3d ago | Python | — |
How does FollowAgents rate this agent?
Why each dimension lost points
Evidence shows least privilege design: session isolation, per-resource permissions, credential brokering, secrets encryption. User confirmation via change requests and approval gates. Data flow transparency with audit trail and session-scoped audit. Sensitive data handling with encryption and injection. Dependency security with many overrides and patches. External effects with approval gates and change requests. Rollback only via git versioning, no dedicated rollback. Source attribution: publisher unverified, but repo has clear commit history. Deductions: rollback and source attribution insufficient evidence.
Self-consistency: README consistent with code structure, tests cover key paths. Dependency availability: uses pnpm and lockfile, but not verified. Failure messages: tests show specific error handling. Deductions: dependency availability unverified, failure messages only partial.
Audience and scenarios: enterprise and developers, self-host and cloud options. Capability boundaries: clearly states agents run in sandbox, changes require approval. Trigger precision: supports cron and webhooks. Environment fit: multiple deployment options. Deductions: capability boundaries and trigger precision only documented, not verified.
Information architecture: clear directory structure. Install notes: quickstart and self-host guide. Naming stability: unified versioning. Examples and FAQ: docs links. Known limitations: not explicitly listed. License: Elastic License 2.0. Versioning changelog: version numbers but no detailed changelog. Maintenance responsibility: publisher unverified. Deductions: known limitations missing, maintenance responsibility unclear.
Output usability: agents return finished deliverables. Marginal value: unique as open-source alternative. Cost benefit: self-host free, cloud priced. Deductions: cost benefit not verified.
Claim traceability: README claims not fully mapped to code. Cross-source corroboration: no independent verification. Fact inference separation: not clearly separated. Deductions: lack of traceability and independent verification.
- Publisher identity unverified, assess supply chain risk carefully.
- Elastic License 2.0 restricts hosted services, may affect commercial use.
- Many dependency overrides and patches, but actual security not verified.
- Rollback relies only on git, no dedicated rollback feature.
- Known limitations not explicitly listed, potential hidden risks.
Recent releases
Latest entries from the project's GitHub Releases. Notes are quoted from upstream as written, not rewritten or tested by FollowAgents.
-
v0.14.2 v0.14.2 — Trigger run history, a bigger Free plan, and steadier sessions
Release notes excerpt
New - Trigger run history. Every trigger keeps a list of its runs with status, error and end time. You can test-fire a trigger with a payload, replay a past run, and missed cron slots are bounded. - More credit tiers and larger top-ups. The Free plan now includes 500 credits and the managed models. - DeepSeek V4.1 Flash now serves the 1017 snapshot on Morph (US region, zero data retention). - Subprocessor notices. The subprocessor list is published, and owners and admins of paid accounts get a 30-day email notice before a new subprocessor takes effect. …View on GitHub ↗ -
v0.14.1 v0.14.1 — Apps work again from older CLIs, members see only their own work, and Capture include lists
Release notes excerpt
New - Kortix Capture records only the apps and websites you include: an organization policy plus a per-computer list. - kortix.yaml version 3 agents can opt in to the project repository. Fixed - kortix apps works again from CLI versions before 0.14.0. Since 0.14.0 every Apps command from an older CLI answered "Apps is not enabled for this project". - A plain member reads and sees only their own work: their own sessions, costs and directory entry. Owners and admins are unchanged. - The Usage cost summary no longer fails on a gateway row with an empty provider. …View on GitHub ↗ -
v0.14.0 v0.14.0 — Credits plan, Enterprise features, restorable sessions and generative UI
Release notes excerpt
New - One credits plan. Every account is on Free or Credits, with top-ups on either. Balances and usage read in credits everywhere; prices you pay stay in your currency. Raising your plan applies at once; lowering it takes effect at the next renewal, with a preview before you confirm. - Enterprise is a feature layer. SSO, SCIM, roles and groups, audit and branding come with an Enterprise license or a self-serve demo. Trials are gone. Admins can now manage billing. …View on GitHub ↗