Æ
g
e
n
t
i
s
k
▍
.
Organisations that work with AI—
and as AI.
Try it. Right now.
Describe your codebase in one line. Get a 60-second sample maturity snapshot — the same shape we deliver in a paid review, just shallower.
Powered by Claude Haiku · this is a sample, not a deliverable. Real review = 6–8 reports, architecture guards, prioritised roadmap.
Five ways to work with us.
Book one — or run the full programme.
What you can get now.
On stage, where the builders are.
javaBin · Agentbrew · JVM, JavaScript and agentic-dev meetups across Oslo and beyond.
Internal events and brown-bags for engineering & product organisations across the Nordics.
Practical, opinionated, grounded in production — not hype.
The 10× Developer Productivity Workshop.
Transform your team into AI-era developers who ship 10× value — with a sharp eye on AI cost.
┌──────┐ ┌──────┐ ┌────────┐
│ build│ → │ test │ → │ change │
└──────┘ └──────┘ └────────┘
│
▼
┌──────┐ ┌──────────┐ ┌──────┐
│repeat│ ← │ improve │ ← │ learn│
└──────┘ └──────────┘ └──────┘
Your own codebase or idea.
A Claude that actually knows what it's working on. Personal license to the Skills Library.
Engineering teams across consulting, product and applied-AI organisations.
- Next cohort · 21 May
- 15 places only · first-come
- Personal license to the Skills Library included
- Bring your own codebase
- On your codebase, on your premises
- Tailored curriculum · paired follow-ups
- Skills Library business license available
- Half-day to multi-day formats
I've been on many AI courses that promise a lot and deliver less. This was different. You show up with your own codebase and work with it all day. Before lunch we had written several skills — rules obvious to me, invisible to any AI.
The instructor clearly doesn't just read about AI — he uses the tools daily on real problems. Claude now orients itself quickly in our codebase. Less noise, and an assistant that actually knows what it's working on. After one day we work noticeably smarter with AI.
Vår visjon i Mynder er nært knyttet til europeisk digital suverenitet. Vi utvikler en ny
infrastruktur som gjør at virksomheter trygt kan ta i bruk AI-agenter — med sikkerheten,
sporbarheten og compliance-forankringen som kreves når det står noe på spill. Tiden for å
eksperimentere er over; nå handler det om å sette agenter i produksjon der de skaper reell
forretningsverdi og kutter kostnader. Det er noe hele Europa etterspør.
Etter at Totto kom inn har vi fått en etterlengtet tempoøkning, og leverer
nå mer og raskere — uten å gå på akkord med kvalitet eller sikkerhet — tvert imot. Han er i
tillegg svært hyggelig og en raus læremester, og det er et privilegium å bygge et agentisk
selskap sammen med ham. Han er rett og slett et unikum, og vi er svært heldige som fikk ham
med på laget.
Make your codebase agent-ready.
- Organisation and team readiness
- Tech stacks and codebase fit for agents
- Which areas can be automated
- How large is the blast radius
- 6–8 technical reports
- Architecture guards
- Security analysis
- Prioritised roadmap for ægentisk adoption
21 deliverables in 7 days.
- D1 scope · index codebase · run synthesis ⌖ kickoff
- D2 architecture analysis · threat model
- D3 security remediation · paired refactor ⌖ midweek review
- D4 frontend + backend fixes · characterisation tests
- D5 strategic planning · cost model · skill packaging
- D6 test pyramid · deploy pipeline · briefings
- D7 handover · demo · train-the-trainer ⌖ sign-off
All figures from git history — not surveys. Team size fell 85%. Engineering output grew 10.7×. Same client, June 2025 – May 2026.
The monthly retainer — five needs.
We build alongside you.
AI digital workers running in production. Human-led, not autonomous. Single feature → full ægentisk transformation.
elprint: MPN normalisation, 17 specialised similarity calculators (resistors, MOSFETs, MCUs, sensors…), BOM validation, metadata for 135+ manufacturers.
24/7 on AWS EC2.
with controlled quality
Memory as infrastructure.
Most AI-dev setups treat memory as text files you maintain by hand. We treat it as a system — and we'll either implement it on your stack, or license you the pieces.
- Lucene-based, sub-second retrieval across tens of thousands of files
- Knowledge graph of directory purpose & archetype
- 21 security signal types — incl. AI-specific threats
- Auto-generates skills from recurring session patterns
Human-led ægentisk workflow. In the loop at the decisions that matter — out of it everywhere else.
$ exo declare intent $ exo execute --dry-run $ exo verify --human
knowledge.yaml → agents understand structure. Saves 33% context.
Agent memory rots.
topic-health + topic-triage run
nightly · zero LLM calls · human in the deletion loop.
Freshness beats completeness.
- Java · Spring · Kotlin · TypeScript · Python · React
- Architecture · security · testing · perf · DX patterns
- Auto-curated from real production sessions
- Behavioural scoring — stale skills retired automatically
- Personal license — included with every workshop seat
- Business license — enquire within
- Updated weekly · audit-trailed · self-hosted option
knowledge.yaml— company, products, team, customers, complianceknowledge.yaml.sig— ED25519 signed, verifiable provenancellms.txt— AI crawler format (llmstxt.org standard)llms-full.txt— full-detail version for agent pipelinesCLAUDE.md— company context hub for Claude / Cursor / ZedAGENTS.md— agent instructions (all frameworks)SETUP.md— deploy to web root in minutes
- Architecture map — directory purposes, patterns, anti-patterns
- Security analysis — 21 signal types, severity rated
- Technical debt inventory — quantified and prioritized
- Modernization roadmap — quick wins + structural changes
- KCP manifest — drop it in, agents understand your codebase
- Auto — fully agent-run, async delivery, from 25 000 NOK / month
- HITL — human in the loop at declare, execute, verify, 125 000 NOK / month
- Works on any codebase — Synthesis indexed, ExoCortex orchestrated
- Deliverable: PR, report, or deployed artifact
- Same workflow used in eXOReaction client engagements
Implementation engagement available — we wire these into your stack and hand it back maintained.
Wake up to a synchronised company.
Mímir composes the pulse. Slack channels carry it. No meetings. No status reports. No copy-pasting between tools.
From the field.
Who builds this.
Four decades shipping production code across Java, Kotlin, TypeScript, Python and the architectures around them. Spent the last few years using agentic tools every day on real client problems — long enough to have opinions, scars, and a working method.
Ægis is the practice that came out of that. The tooling (Synthesis · ExoCortex · KCP), the workshops, the case studies — all of it is what I do daily, packaged so a team can adopt it without re-living the learning curve.
From zero to ægentisk — in six weeks.
Infra cost ~15,000 NOK/month. Positive ROI the moment each team member saves 20 minutes a day.
Network isolation · financial outputs cite API sources · sensitive queries DM-only · hard spend alerts · documented backup admin · gradual channel rollout.
From scoping call to self-sufficient team.
Or pick any single offering — talk · workshop · review · build · tooling.