Docmancer is built around two questions, in this order:
- I re-explain my project to every coding agent. How do I get them all working from the same context?
- What do my coding agents already know about me and my working style?
It discovers memory, instructions, rules, and eligible session evidence that supported coding agents already wrote, keeps the durable parts as one canonical Markdown tree, and delivers the relevant files back to every agent you connect. Source files remain untouched, and documentation remains a separate Library corpus.
Two further problems sit outside what a single machine can solve on its own, and those are what the paid tiers exist for: getting the same memory onto your other machines, which is Personal Sync, and sharing approved context with colleagues, which is Team Sync and is not available yet.
Install and discover
pipx install docmancer
pipx ensurepath
export PATH="$(pipx environment --value PIPX_BIN_DIR):$PATH"
docmancer setup
The PATH export makes Docmancer available in the current terminal, including an existing SSH session. pipx ensurepath keeps it available in future terminals.
If you use uv, uv tool install docmancer installs it the same way.
setup shows one preflight plan and a privacy warning, then waits for confirmation. Once you confirm, it indexes existing agent memory, builds the machine-wide Markdown tree under ~/.docmancer/tree, installs supported user-level integrations, and enables recall and lifecycle capture for the agents that support them. It does not change the project in your current directory.
The default retrieval stack needs no daemon or model download. Pass --profile scale instead when you want the Qdrant and FastEmbed stack:
docmancer setup --profile local
docmancer setup --profile scale
Open the human interface
cd /path/to/project
docmancer web
- Home is for grounded Ask, customising Docmancer, and connecting coding agents.
- Shared Memory shows every canonical file, its folder arrangement, and what connected agents receive.
- Library separates curated memory, agent evidence, and technical documentation.
- Settings chooses a generation provider and model, stores credentials in the operating-system keyring, and manages capture and optional Cloud.
Ask what is already known
docmancer ask "Why did we choose Railway?"
docmancer ask "What changed in the release process?" --no-answer
Recall works without a generation provider. When a provider is configured and ready, Ask calls it after retrieval and returns grounded prose with citations and verification checks. Use --no-answer for the cited evidence bundle only, and --fresh when the question must wait for changed agent files to be indexed first.
Browse the memory scaffold
Shared Memory is ordinary Markdown arranged into two trees. Machine-wide memory uses profile/, principles/, projects/, and shared/. Project memory uses decisions/, constraints/, workflows/, and lessons/.
Open any file to read its body, source lineage, scope, authority, and stable address. Select a connected agent to preview the exact bounded projection it receives. This path does not require an LLM.
Connect an agent explicitly
docmancer agent install claude-code --hooks
docmancer agent install codex --hooks
Detection, skill installation, managed instructions, automatic recall, capture, and recent successful use are different states, and Shared Memory reports them separately. --hooks installs recall on its own, and --capture-hooks installs lifecycle capture on its own, so you can add or remove either without touching the other. Claude Desktop requires a manual upload of the generated skill package.
Keep a deliberate decision
docmancer write $'# Release process\n\nDeploy the API on Railway.' \
--path decisions/release.md \
--scope project
Use docmancer read, edit, and move for subsequent changes. Existing-file mutations require the current content hash.
Continue on this machine or add another
At this point the complete local product is ready. Run docmancer status whenever you want one check across memory, delivery, security, retrieval, and optional Cloud health.
If this is the only machine you use, there is nothing else to buy or enable. If you want the same canonical memory on another machine, follow the Personal Sync guide. Personal Sync is optional and paid because it adds encrypted transport, managed revision history, and recovery outside your machine. It does not improve or unlock local recall.
docmancer cloud connect
The connection creates and checks a recovery kit, then starts the first encrypted sync automatically. On another machine, use the four-word pairing code or your recovery kit.
Team Sync, which will carry approved context to colleagues, is not available yet.