Simple
Run projects without project-management bureaucracy. No epics, story points, velocity, Gantt charts or workflow builders. If it needs training to understand, we reconsider it.
Calm software, by design rule
Projects, conversations, decisions, knowledge and AI agents —running on infrastructure you control.
docker compose up -d runs on a modest VM 2 vCPU / 4 GB no telemetry by default
Project · Platform Rebuild
Alex· 14:31
Leaning PostgreSQL for the transactional store. Consistency matters more than raw write speed here.
Developer Agent· 14:36read-only scope
Agreed. Benchmarks in the docs folder confirm p95 under 8 ms at 10k rows/s. Proposing we record this as a decision.
AI proposal · extract decision
We appear to have reached a decision.
Built for companies of 10–100 people
The core loop
The key concept isn’t “project management.” It is a loop where every conversation has an outcome — and both humans and agents participate.
Long-form asynchronous discussions that create real organizational records — not noise.
Every discussion can end in a first-class decision: what, why, alternatives, who decided.
Decisions become curated company knowledge — versioned, sourced, linked, verifiable.
Knowledge becomes context. Humans and agents do better work, and the loop starts again.
Projects eventually finish. Tasks disappear. Chat becomes noise.
But decisions + knowledge + history increase in value over time.
Why TaskLodge
Run projects without project-management bureaucracy. No epics, story points, velocity, Gantt charts or workflow builders. If it needs training to understand, we reconsider it.
Calm software, by design rule
Your infrastructure. Your data. Self-hosted on a Docker VM, with your own object storage, your own database, and AI traffic that never routes through our cloud.
Sovereignty is the default, not an add-on
Give AI workers identities, permissions and access to exactly the knowledge they need. Agents are users — not plugins — with scoped tokens, approvals and a full audit trail.
The office, not the employee’s brain
Five jobs, one workspace
One understandable place for projects, tasks, discussions and documents.
Don’t let important decisions disappear inside chat history.
Turn operational knowledge into something the company reuses.
Identities, permissions and limited access for AI employees.
Run everything on infrastructure you control.
The killer feature
Most collaboration systems accumulate information but never convert operational discussion into durable knowledge. TaskLodge does. Inside any discussion, one action —Extract decision — and AI drafts the record. A human approves it.
AI proposes knowledge. Humans promote it into organizational truth.
Decision · D-0142
AcceptedAugust 27, 2026 · Platform Rebuild
Related Architecture / Databases
Revisit August 2027
Company knowledge
git syncEngineering
Sales
Operations
$ tasklodge knowledge sync
→ pushed 4 files to /company · history preserved · diffs available
Organizational memory
Knowledge is organized in a hierarchy — Engineering, Sales, Operations, Company — and every article keeps its author, contributors, version history, source discussions and related decisions.
Knowledge mirrors to a Git repository for portability, history and agent access — while staying usable for someone who has never used Git.
“This policy hasn’t been reviewed in 14 months” is far more valuable than another “Ask AI” button.
Projects, decisions, knowledge, tasks, people and agents form one company graph — so you can ask why, not just where.
Agents
An agent is represented like a human principal: identity, role, permissions, knowledge scope and an approval policy. Every significant action is attributable.
Role · Market Research|Projects · Product Strategy, Competitor Research
Can
Cannot
Approval required
Attribution · Research Agent
Aug 27 — 14:42
Bring your own runtime
We don’t ship an agent runtime. TaskLodge is the office, not the employee’s brain.
Self-hosting
Installation, upgrades and backups receive the same design attention as task creation. The deployment target is a modest VM — and one line.
$ docker compose up -d
[+] Running 5/5 app · postgres · minio · worker · caddy
$ tasklodge backup
✓ backup-2026-08-27.tar.gz
$ tasklodge restore backup-2026-08-27.tar.gz
✓ restore verified automatically
# or, when we ship the installer:
$curl -fsSL https://tasklodge.example/install | sh
A modular monolith behind a reverse proxy: app, PostgreSQL, object storage, background worker. No Kubernetes, no microservices.
MinIO, AWS S3, Cloudflare R2 or any compatible system. Your files live where you point them.
tasklodge backup and tasklodge restore, covering database, files, configuration and knowledge — with automatic restore testing.
Point TaskLodge at OpenAI, Anthropic, Azure, a local model or any OpenAI-compatible endpoint. Your AI traffic never routes through our cloud.
Self-hosted installations ship with no customer-data telemetry. Anonymous product telemetry is opt-in — and agent prompts stay yours.
Exporting company data is straightforward. Lock-in should come from years of valuable memory, not from fear of getting your data out.
Pricing
Your infrastructure. Your users. Your agents. No per-seat math on 43 employees plus 27 agents — one price for the instance.
Freeforever
A genuinely complete core. Our distribution engine — run it as much as you like.
$990/year per instance
For companies of up to 25 humans — with unlimited agents.
$2,990/year per instance
For companies of up to 100 humans that need governance.
$12,000/year · from
Air-gapped, compliant and supported — quoted to your requirements.
Target prices from our bootstrap plan — final pricing will be tested with design partners. Managed cloud editions come later; self-hosting stays the brand.
Why now
“Who controls our collaboration infrastructure? Which applications must be cloud? Which systems should remain sovereign?”
Design partners · 5 seats
We are recruiting five companies of 10–100 people to run at least one real project on TaskLodge and give us structured feedback. If the loop doesn’t hold, we’d rather know now.