ADHED — self-hosted task + notes backend as a NanoClaw skill

[dimatosj]

Built a headless task management API and open-sourced it with a NanoClaw integration included.

What it is: A self-hosted REST API (FastAPI + Postgres) for task management and knowledge fragments. No UI — API only. Designed as a shared backend that agents, CLIs, and scripts all write to, with a server-side rules engine that enforces workflow for everyone.

The NanoClaw skill: Ships in the repo at integrations/nanoclaw/life/. It's a single-file Python CLI (life-cli.py, stdlib only) that wraps the full API. Covers issues, projects, knowledge fragments, rules, notifications, triage, and bulk operations. Drop it in your container's skills directory and set TASKSTORE_URL, TASKSTORE_API_KEY, TASKSTORE_USER_ID, and TASKSTORE_TEAM_ID as env vars.

Fragments are the interesting part for agent workflows — typed notes (person, place, credential, memory, idea, resource, journal) with topics, domains, entities, and full-text search. Your agent captures context from conversations and stores it as structured fragments that it can query later.

Rules engine runs server-side on every write. Examples: reject creating tasks without a priority, auto-label based on title keywords, limit work-in-progress, notify on assignment. Write a rule once and every client gets enforcement for free.

Setup: git clone + ./setup.sh (needs Docker). Takes about 30 seconds.

Repo: https://github.com/dimatosj/adhed
Integration docs: https://github.com/dimatosj/adhed/tree/main/integrations/nanoclaw