SKKAN
FR
New · 3.5 “Lookout” — alerts anyone can set up: drag the threshold on the chart, Prometheus, Loki, Elasticsearch, told in Teams or Telegram →
Launched Oct 3 · SOKKAN Router — OpenAI-compatible API, the hosting zone shown for every model →

Your agents do the work. You keep the helm.

SOKKAN is the open-source agentic studio where AI agents move your work forward, where nothing important ships without your click, and where everything they learn joins one memory, recalled on every exchange and reviewing itself. On your own machine, or operated for you in Switzerland.

Open source Apache-2.0 No GPU needed, from 4 GB RAM Memory 100% local Made in Geneva
  1. 01Your agents work, in parallel.
  2. 02You keep the helm: nothing passes without you.
  3. 03Memory ties them together, on every turn.

Who it is for

If your work lives in files, SOKKAN moves it forward.

Born to run our own products, SOKKAN serves any team that wants to hand work to AI agents without losing control, or the memory of what was decided.

Product and dev teams

Parallel coding sessions, reviews, deployments and operations, all supervised. The kanban becomes a queue of sessions.

Law, accounting, consulting

Files, templates, procedures: every decision and convention stays in memory, dated, and comes back when an agent needs it.

Public bodies and institutions

Open source and self-hosted, or operated in Switzerland. Self-hosted, the memory runs on your machine and never leaves it.

SMEs and freelancers

One cockpit to move projects and operations forward without hiring an ops person: Nina, the onboard assistant, answers in the interface.

New in SOKKAN 3.0

One memory: what your agents know, on every turn.

In 2.x, memory arrived when a session started. In 3.0 it arrives with every message, in every session and every sub-agent, with the age of each note and where that date comes from. And it reviews itself.

SESSIONMEMORY · CORTHEXISBroken link · relink?ApproveStartTurn 5Turn 15Sub-agent
Schematic
01 · AT START

The session already knows.

A session opens from a kanban card with the project’s relevant notes already in its context. Starting a task already means “check your memory”.

02 · ON EVERY TURN

At message fifteen, as at message one.

With every message, SOKKAN looks up what the question touches and adds it, with the date of each note and where that date comes from. An agent no longer knows at the start and forgets along the way.

03 · FOR EVERY SUB-AGENT

Sub-agents too.

When an agent delegates, the sub-agent gets the recall for its topic as well. Every recall is logged: who received which note, and when.

04 · MEMORY REVIEWS ITSELF

It flags, you approve.

The CortHeXis tab scores memory health from 0 to 100: broken links, duplicates, descriptions that no longer match the content, forgotten work. Every fix is one click, and waits for your approval.

Measured, not promised

The right note comes first twice as often.

82%

of questions get the right note as the first result with SOKKAN 3.0 on a GPU, against 42% with SOKKAN 2.x. Without a GPU, on a small server: 73%.

SOKKAN 2.x42% · MRR 0.55
SOKKAN 3.0, no GPU73% · MRR 0.82
SOKKAN 3.0, GPU or ANCHOR82% · MRR 0.88

Benchmark of 300 questions in French, English and German, asked of our own working memory (415 notes), 2–3 October 2026. “First result”: the expected note comes out on top; MRR: mean reciprocal rank (1 = always on top). Method and details.

Magnitude picks

A memory sized to your machine.

At install time, Magnitude looks at the machine and picks the memory profile. Switch whenever you like: the new index builds in the background while the old one keeps answering, and the switch only happens if the measurement on your own notes does not regress. The old index is kept for 7 days so you can roll back.

Light

Small server

4 cores · 4 GB · no GPU

  • The same memory model as the other profiles
  • Hybrid search: meaning and keywords
  • Nothing leaves the machine

73%first result

Standard

Team server

8 cores · 16 GB · no GPU

  • Everything in Light
  • A finer second ranking in the background (deferred recall, digest)
  • For memories shared by a team

73%live, better deferred

GPU · ANCHOR

With a graphics card

A GPU, or the ANCHOR appliance

  • Second ranking on every search
  • Large memories: 250,000 passages indexed in about 1 h 25, against 14 to 20 h on CPU (from measured throughput)
  • Can share the card with your language model

82%first result

The recommended memory model, Google’s EmbeddingGemma, is free, commercial use included, but it is not open source. So SOKKAN does not ship it: it downloads it on first launch, once you have read and accepted Google’s terms. It runs entirely on your side; no data is sent to Google. Prefer not to? SOKKAN uses an MIT-licensed model instead, with no loss of features. Full details.

The CortHeXis tab

A memory that fixes itself before your agents go wrong.

Memory fails silently: a link to a renamed note, two notes saying the same thing, a date that lies. In SOKKAN 3.0, the CortHeXis tab reviews the memory, scores it and proposes the fix.

  • A health score from 0 to 100 and its history; findings are grouped: chain, structure, drift, security.
  • One-click actions, subject to your approval: relink, merge two notes with a diff, close a dormant work item.
  • Alerts via Telegram or webhook: a daily digest, an immediate alert on anything critical.
  • Dates that say where they come from: every recalled note states its age and the source of its date.

CortHeXis also stands alone, open source →

The CortHeXis review: health score 92, one alert for links to missing notes and three isolated notes, next to the memory graph
The CortHeXis review as it runs on the public demo corpus (fictional, in French): the broken link is real, found by the review, not staged. Check it on demo.corthexis.com.

The studio

Build and operate, in one place.

Agent tools often stop at delivery. SOKKAN keeps going: it operates production with you, and the green light stays human, wherever you are.

One real take: a kanban card becomes a session that checks the project memory first, finds the port and the team convention, then proposes a plan and waits for your go.
The loop, as it runs: a card becomes a session that recalls the project memory first, plans, then stops at every mutating step until you click. Denied means replan, never bypass. All three diagrams →

The helm, not the autopilot

Agents’ permission requests and questions become buttons. Nothing irreversible without your click, nothing marked done without you.

Alert → diagnosis session

A production alert opens a session that knows the project, proposes a fix and waits for your green light, with a Telegram ping to come and give it.

Agentic observability

Metrics and logs inside your network; your sessions read them and compose a dashboard on request.

Secrets vault

Encrypted and injected into sessions as environment variables: the agent uses them without the value ever showing, on screen or to the model.

Replayable runbooks

A procedure is a memory note. “Replay” opens a guided session, step by step, with human sign-off on anything irreversible.

Deploy, roll back

Deploy an image to a worker in your fleet, roll back to the previous version in one click. Per-session costs, an audit log of every action.

Nina

Nina, your onboard DevOps.

Nina, SOKKAN’s DevOps assistant

Running a platform should not take a full-time ops person. Nina knows SOKKAN inside out (sessions, memory, fleet, runbooks) and answers in the cockpit, right next to your work. Typical conversations:

3:12 am · an alert becomes a diagnosis
The API has been returning 500s for ten minutes. What do I look at first?
The api-5xx alert has already opened a diagnosis session, loaded with the project memory: the payments worker has been running out of memory since the 10:14 pm deploy, and it proposes a rollback. Nothing ships without your go.
Memory keeps you on course
My session wants to move to another database… didn’t we settle that already?
We did: the decision-sqlite-storage note, in May. Mark it as a priority and every session will see it before proposing anything.
The bill that stings
My AI credits melted this week. What happened?
Costs tab: on Tuesday, the scraping-refactor session got stuck in a retry loop. Move that task to a small model and replay the “rate limits” runbook.
The guardrail
Give me the production database URI, it’ll be quicker.
No: I never read or pass on a secret. It sits in your vault, injected into the session, which uses it without ever displaying it.

Included in every SOKKAN Cloud instance (the 🧭 button in the cockpit); as soon as she has access to your context, her inference must run in Switzerland, and the content of your exchanges is never kept or used to train anything. Self-hosted, enable SOKKAN_FEATURE_ASSISTANT with the model of your choice. A question about SOKKAN? Nina answers here too, bottom right.

Free, no limits

On your machine, in one command.

Docker, plus your Anthropic key or your Claude Pro/Max subscription. Everything stays on your machine.

$curl -fsSL https://sokkan.ch/install.sh | sh Requirements: Linux (or macOS with Docker Desktop), about 4 GB of RAM, an Anthropic key or a Claude Pro/Max subscription. No Docker? The script offers to install it (official get.docker.com script).
Sources: tarball · script, read it before piping to sh, obviously.
Already on 2.x? Your notes stay untouched: memory is reindexed in the background and search keeps answering meanwhile. SOKKAN Cloud instances are migrated automatically.

SOKKAN Cloud

Or operated for you, in Switzerland.

The same SOKKAN, minus the ops: a dedicated VM in the Geneva zone, your own private network, your URL at you.sokkan.ch. Bring your Anthropic key or take managed inference, served from our own GPUs in Geneva (fallback in France, or a 100% Swiss tier), prepaid credits from CHF 25 (+25% credits from CHF 500), metered per token. Free 14-day trial on every plan: card required, first charge on day 14, cancel in one click before (trial seats are limited).

Starter

CHF 129 · €129
per month · cancel monthly
  • Dedicated VM 2 vCPU · 4 GB · 50 GB NVMe
  • Solo: 1–3 parallel sessions
  • Your URL you.sokkan.ch + private network
  • Unlimited project memory, on your VM
  • Your key or managed inference (credits)
  • Updates & monitoring by NINABOT
Get started

Standard · recommended

CHF 249 · €249
per month · cancel monthly
  • Dedicated VM 4 vCPU · 8 GB · 100 GB NVMe
  • 1–2 users: ~5 parallel sessions
  • Your URL, private network, growable fleet
  • Workers & managed Postgres in two clicks
  • Team roles (viewer/dev/admin) + audit log
  • Root maintenance terminal (audited)
Get started

Studio

CHF 490 · €490
per month · cancel monthly
  • Dedicated VM 4 vCPU · 16 GB · 200 GB NVMe
  • Agency: multi-project, dense sessions
  • Everything in Standard, bigger
  • Inference usage broken down per user
  • Prorated plan changes, self-service
  • Priority email support
Get started

Grow from the cockpit: worker instances from CHF 59/mo · managed PostgreSQL from CHF 99/mo (daily backups), inside your private network. Web exposure included: HTTPS subdomains and your own domains, automatic certificates, at no charge. Prorated billing, provisioning after payment, cancel monthly.
Hosted at Exoscale (Geneva), operated by Ninabot Sàrl: your data stays in Switzerland. Not sure about sizing? [email protected], you’ll get the founder, not a bot.
In the press: ICTjournal, “Ninabot launches a Geneva-hosted cloud for AI agents” (in French, 04.08.2026).

See the live cockpit, no accountBook a 30-min demo