The leading open-source alternative toClaudeClaude CoworkandOpenAIChatGPT Work

The open-source AI Management System

Your agents, their skills, your company memory and every connector in one platform. Any model, your keys, self-hosted or managed cloud.

Open source · Any model, your keys · Self-host, VPC, or on-prem

Runs any model
ClaudeOpenAIGeminiOpenRouter
3,000+ apps, any MCP or API
Github
The Kortix stack

Every layer an AI workforce needs — in one platform you own.

Most tools hand you one layer and rent you the rest. Kortix is all six — they read in order, and the sixth commits the work back into the first.

01

One git repo that is the company.

Agents, skills, memory, connector config and triggers are all text in one repo. Not settings in someone else’s database — files you own.

  • kortix.yaml declares the machine image, the connectors and the triggers
  • Agents and skills are markdown; memory is files that accumulate
  • grep the whole company, diff any change, roll any part of it back
acme-coyour whole company, as files
Agentswho does the work
Skillshow your company does a job
Memorywhat it has learned so far
Ruleswhat each agent may touch
Triggerswhen work starts on its own
Connectorsthe tools it is wired to
Machinethe image sessions boot
02

Every tool your company already runs on.

The repo also holds the connectors. Wire up Slack, docs, tickets, CRM, billing and code once, then scope which agent may touch which one.

  • 3,000+ apps in a click, plus MCP, OpenAPI, Postman, GraphQL and raw HTTP
  • Connector credentials are brokered server-side and never enter the machine
  • Rule each tool call allow, ask or block — down to the arguments it was given
03

Run any model. Keep your keys.

Kortix is model-agnostic. Pick the model per agent, per session or per message, and switch the day a better one lands.

  • Bring your own API key from any major provider, or use ours
  • Or sign in with the ChatGPT subscription you already pay for
  • Or your own models, behind your own URL — anything OpenAI-compatible
Claudemodel: autoper agent · per session · per message
ClaudeAnthropicOpus · Sonnet · Haiku
OpenAIOpenAIGPT · Codex
GeminiGoogleGemini Pro · Flash
Your endpointanything OpenAI-compatible
04

The part that turns a model into an agent.

A model on its own answers. The harness gives it planning, tool use and multi-step runs it actually finishes — powered by OpenCode, configured by a file in the repo.

  • An OpenCode agent: markdown, plus the tools and plugins beside it
  • Say allow, ask or deny per tool — down to a single shell command
  • Open source, so the harness is never the thing you are locked into
With the harnesspowered by OpenCode
  • Read the webhook handlerwebhook-retry.ts
  • Patched the retry to back off+18 −6
  • Ran the billing suite3 passed
  • git push origin mainblocked by the agent file
  • Opened a change requestfor a human to read

agents/kortix.md · permission.bash → git push: deny

05

Every session gets its own computer.

The harness needs somewhere to run. Each session boots its own isolated Linux machine with your repo and your tools already on it — nothing to install, no laptop involved.

  • Session id, sandbox id and branch name are one and the same string
  • The agent can install, run and break anything — only commits survive
  • Thousands run in parallel with no crossover between them
06

One place to start it, one gate to land it.

Start, watch and steer every agent in the organization — from the web app, from Slack, from your terminal, or from nobody at all. Then the work commits back into the repo you started from.

  • Web, Slack, mobile, CLI and API all start the same session
  • Cron schedules and signed webhooks start sessions with no one asking
  • Session work lands on main as a change request you read as a diff first
By role

It picks up work from every team and runs it start to finish.

Outbound, error triage, the books, the Monday report — real jobs, on real machines, landing as changes you can read.

Sales01

Builds the outbound campaign and runs it.

It pulls a lead list from Apollo, enriches every account, and writes a sequence per lead. Put Ask on the send step and it stops with you before anything goes out.

outbound/northwind-01.emlOutput
ops@northwind.example08:12

Your four new platform hires

Saw you are hiring four platform engineers.

Teams at that stage lose about a week a month to

manual reconciliation. We built an agent that closes

that loop and shows its working. Worth twenty minutes?

Draft · Ask set on send

240 leads pulled · 24 enriched · 3 sequences written
Engineering02

Reads the logs, ships the fix.

Every morning it sweeps the day’s errors, groups them, reproduces the worst one on its own machine, patches it and opens a change request against main.

fix-checkout-expiry.diffOutput
app/checkout/charge.ts
if (card.expMonth < now.getMonth()) {
if (isExpired(card, now)) {
return fail(402, "card_expired");
}

+2 −1

9 error groups · top one fixed · 118 tests pass
Marketing03

Runs the SEO programme week after week.

It tracks the queries you care about, finds the pages losing ground, rewrites them against the brief, and opens each rewrite as a change request.

marketing/seo-weekly.xlsxOutput
QueryPagePosChange
payout reconciliation/guides/payouts4+7
invoice chasing/features/ar9+3
bank feed sync/guides/bank-feed14−2
month end close/guides/close2+1
receipt capture/features/receipts21new
5 briefs written · 2 rewrites in a change request
Finance04

Does the books, not just the summary.

It reconciles the ledger against the bank, chases the receipts that are missing, attaches them, and closes the month with the variance explained.

finance/fy26-budget.xlsxOutput
Cost centreH1 planActualVariance
R&D — platform1,225,0001,182,400+42,600
Go-to-market810,000742,800+67,200
Security & compliance180,000198,600−18,600
Brand & content240,000219,400+20,600
Total2,455,0002,343,200+111,800
9 receipts chased and attached · 1 centre over plan
People05

Sources the pipeline, not just the inbox.

It searches the sources you connect for people matching the brief, scores each against the written bar, and drafts the first approach for every one worth it.

hiring/staff-infra-pipeline.xlsxOutput
ProspectSignalScoreStage
#118Ran platform at 200+91replied
#094Infra lead, six years84contacted
#131Strong, wrong stack67sourced
#072No systems depth38passed
340 profiles searched · 22 shortlisted · 9 approached
Product06

Turns what users say into what ships.

It reads every piece of feedback that came in, groups it into themes you can act on, and takes the top one all the way to a change request.

product/feedback-2026-w31.mdOutput
Export times out over 10k rows14
Wants CSV as well as xlsx9
Filter state lost on reload7
Onboarding called out as good12
4 themes · top one fixed in a change request
Data Science07

Has Monday’s numbers before Monday.

It queries the warehouse on a schedule, checks the result against last week, draws the chart and posts the whole thing to Slack while you are asleep.

reports/win-rate-weekly.mdOutput

Win rate by quarter

Q3 2536.7%
Q4 2544.7%
Q1 2654.2%
Q2 2666.7%
Posted to #revenue · top loss reason: price
—

GitHub stars on kortix-ai/suna

We are building the open AGI platform. Every company should own all of it.

Security & trust

Giving agents real access is the easy part. Trusting them with it is the work.

How Kortix is secured
  • In progress
  • In progress
Frequently asked questions

The questions people ask before the first session.

Run your whole company from one repo you own.

Start with one job and grow from there.