Governance platform for AI agents

A sovereign cloud workspace for AI agents.

Code in your browser, on your infrastructure. Hubs that ship. BrainDB that remembers. Every commit, decision, and memory write gets a receipt. On-prem or our EU cloud.

Request enterprise access See it for solo + small teams

audit-trail.live

live

TimeWho → did whatReceipt

--:--:--agent:coding→code.commit#a8f2
--:--:--user:mitchell→tenant.provision#b14c
--:--:--agent:brain→memory.write#c903
--:--:--agent:compass→decision.record#d6e1
--:--:--user:dean→settings.update#e7a8
--:--:--agent:sales→artifact.export#f211
--:--:--agent:coding→code.review#a45d
--:--$

Per-tenant isolation · append-only7 of ∞

The platform

Build, ship, and govern AI agents on one stack.

AVAILABLE STANDALONE

Academy

Compass

Assistant

Legal

Sales

Coding

orchestrating 6 domain hubs

Hubs

Per-domain engines that ship code, training, and decisions. One contract, six hubs today.

Explore Hubs

POST /api/v1/memory
{
  namespace: "brain/decisions/...",
  value: { ... },
  rationale: "...",
  owner: "user:m",
  dependencies: [],
}▍

BrainDB

Governed memory. Every write carries rationale, owner, dependencies. Append-only audit.

Explore BrainDB

openvaluehealthy

demo-3healthy

techrebelsrefreshing

larshealthy

+ 70 tenants

Control Tower

Operator dashboard for the tenant fleet. Provision, refresh, monitor, audit.

Explore Control Tower

14:02user:mitchell → tenant.provision

14:05agent:coding → code.commit

14:11user:dean → settings.update

14:18agent:brain → memory.write

14:23user:mitchell → tenant.delete

Audit

Every decision recorded. Per-tenant isolation. Queryable by actor, action, tenant, time.

Explore Audit

BrainDB

Memory you can navigate.

Every decision carries rationale, owner, and dependencies. Semantic recall surfaces past decisions when you need them. No re-litigation.

brain/decisions/use-pgvector

decisions

Mar 12 14:23user:mitchell→decision.record#d9c1
“Picked pgvector over Weaviate. Lower op cost. Same recall on our embedding model.”
Mar 14 09:11agent:coding→code.commit#a8f2
“Migrated brain-storage to pgvector per decision #d9c1.”
Mar 28 16:42agent:brain→memory.query#e4b3
“Semantic recall of #d9c1 at 0.94 similarity — surfaced when re-evaluating vector DB choice 2 weeks later.”
Apr 02 11:08user:dean→decision.confirm#f902
“Confirmed pgvector still right call. Rationale carried forward. No re-litigation.”

rationale preserved · 47 days · 1 recall · 0 re-litigations

brain-storage.ts

+3-4typescript

1// brain-storage.ts

2import { WeaviateClient } from "weaviate-ts-client";

3import { Pool } from "pg";

5const client = new WeaviateClient({

6 host: env.WEAVIATE_HOST,

7});

8const pool = new Pool({ connectionString: env.DATABASE_URL });

9// Decision: brain/decisions/use-pgvector:v1 (rationale=#d9c1)

commit landed under decision #d9c1 · receipt #a8f2

Architecture

Six hubs. One BrainDB. Every action audited.

Each hub ships its domain — code, decisions, assistance, content, contracts, sales artifacts. Every action emits an audit event into BrainDB. Tenants stay isolated; governance scales by construction.

tenant-architecture.svg

live

Per-tenant isolation · audit-on-by-defaultv1.4 governance

shipHubs ship code, content, decisions. Each speaks the same HubInput → HubOutput contract.

decideBrainDB stores the why. Owner, rationale, dependencies — governance baked into the write path.

auditAppend-only audit log on every cross-hub interaction. Replay any decision, any time.

isolatePer-tenant networks + brain instances + storage. Zero cross-tenant data flow by construction.

Operations

Watch every agent task move through governance.

Pending approval, running, done, failed. Every state transition is recorded with rationale, owner, and dependencies. No silent moves.

control-tower / audit-board

live

Pending approval2

wo-014

agent:coding→code.deploy

Touches auth middleware — awaiting operator gate

wo-015

agent:sales→artifact.publish

Customer-facing deck — pending Legal review

Running3

wo-012

agent:compass→decision.record

wo-013

agent:coding→code.refactor

wo-016

agent:brain→memory.embed

Done4

wo-008#a8f2

agent:coding→code.commit

wo-009#b14c

user:mitchell→policy.approve

wo-010#c903

agent:assistant→context.compress

wo-011#d6e1

agent:brain→memory.write

Failed1

wo-007#f211

agent:sales→artifact.export

Quota exceeded — retry queued

10 tasks · last update 12s agofilter: tenant=openvalue

Governance Loop

Four steps. One trail. Continuous.

Capture every action. Govern the write path. Recall what was decided. Audit forensically. This is the loop that runs by construction.

Capture
01
Govern
02
Recall
03
Audit
04

Agents act, your trail records
Every commit, decision, and memory write emits an AuditEvent — actor, action, tenant, timestamp, receipt. Nothing slips through.
Rationale at the write path
Every BrainDB write requires rationale, owner, dependencies — the three governance fields baked into storage. No rationale, no write.
Semantic search over the why
pgvector-backed recall finds the decision you already made. No re-litigation when context resurfaces months later.
Append-only, forensically queryable
Per-tenant audit isolation. Queryable by actor, action, tenant, time. Designed for HIPAA-eligible review.

Sovereignty

Your cloud. Your boundary. Audited across regions.

Per-tenant deployments span where your data lives. Audit events propagate across regions on the same governance trail.

Seen enough?

Stop guessing which agent did what. Govern it.

Request access Book a call

No demo gauntlet — a founder answers within a day.

5.0vs. tools your devs already use

Your devs keep their tools. You get a paper trail.

Odin layers governance on top of Cursor + Copilot. Same agent output your devs already produce, every commit attributable.

capability	Cursor	Copilot	+ Odin
AI-driven code edits	✓	✓	Hubs ship code
Multi-file refactors	✓	◯	Codex variants
Audit trail of every action	—	—	Append-only
Per-decision rationale	—	—	BrainDB owner+rationale
Per-tenant isolation	—	—	Network + storage
Designed for HIPAA-eligible	—	—	On-prem option
✓ provided · ◯ partial / unsupervised · — not provided

AI-driven code edits

Cursor✓

Copilot✓

+ Odin✓Hubs ship code

Multi-file refactors

Cursor✓

Copilot◯

+ Odin✓Codex variants

Audit trail of every action

Cursor—

Copilot—

+ Odin✓Append-only

Per-decision rationale

Cursor—

Copilot—

+ Odin✓BrainDB owner+rationale

Per-tenant isolation

Cursor—

Copilot—

+ Odin✓Network + storage

Designed for HIPAA-eligible

Cursor—

Copilot—

+ Odin✓On-prem option

✓ provided · ◯ partial / unsupervised · — not provided

Built for Your Organization

Whether you are a growing team or a regulated enterprise, ODIN adapts to your scale and compliance requirements.

Compliance-first

Regulated Industries

Healthcare, legal, finance, government. ODIN runs on your servers with full audit trails, designed for environments where data sovereignty is non-negotiable.

Your infra

SEALED

Your data

HIPAA-eligibleSOC 2GDPRFull audit trail

5–50people5-50 people

Growing Teams

Stop losing context when people leave. ODIN captures decisions, code rationale, and institutional knowledge so your team builds on what came before.

50–500people50-500 people

Mid-Size Organizations

Align departments around a single source of truth. The Brain connects engineering, legal, sales, and operations into one queryable knowledge layer.

Ready to Deploy AI Your Way?

Odin runs on your infrastructure, in your jurisdiction. Three paths in — pick the one that matches your compliance and scale requirements.

Healthcare

Built for Clinical Environments

Designed for HIPAA-eligible deployments. Your patient data and institutional knowledge stay inside your own infrastructure — no third-party data routing.

Request access for healthcare teams

Audit stream — your infrastructure

08:14chart.readon-prem

08:21note.writeon-prem

08:47memory.writeon-prem

09:02audit.emiton-prem

Legal

Designed for Legal Teams

Designed for SOC 2 readiness. Client matter data, contract context, and privileged communications remain in your sovereign environment with a full audit trail.

Request access for legal teams

Enterprise

Governed AI for the Fleet

Multi-tenant fleet management, per-department isolation, and operator-level visibility into every agent action. Priority support included.

Talk to our enterprise team

From the Founders

Why we built AI that organizations own, not rent.

Mitchell Tieleman

CTO & Founder

Dean Falix

CEO

NLDutch Company

GDPR Native

EU Hosted

Get started

Put a name on every action your agents take.

Two paths in. Pick the one that fits your environment.

Request enterprise access See it for solo + small teams

Rather just talk it through? Book a call — a founder answers

capability

Cursor

Copilot

+ Odin

AI-driven code edits

✓

Hubs ship code

Multi-file refactors

✓

◯

Codex variants

Audit trail of every action

—

Append-only

Per-decision rationale

—

BrainDB owner+rationale

Per-tenant isolation

—

Network + storage

Designed for HIPAA-eligible

—

On-prem option

✓ provided · ◯ partial / unsupervised · — not provided