projectdocs and the AI

The AI in the chat panel reads and writes projectdocs through the projectdocs_* tools of the Tracemoor MCP server. Reading works in every session mode. Tools that write need Brave mode. See Session modes.

What the AI does with projectdocs

Tracemoor gives the AI these working rules with every session:

For questions about history, reasons or open work, search the projectdocs before the code.
Before changing a symbol, look up the entries and specs anchored to it. If a change contradicts an anchored decision, say so instead of overriding it.
After writing about code in an entry, anchor the central symbols. After a commit, add a commit anchor.
Bug reports: one file, status open → in-progress → closed, the fix as a Fix (date) section.

The rules are instructions to the AI, not checks. Review what the AI wrote in the dashboard.

Start a chat from an entry

An entry can be the starting point of a new chat session, for example a plan the next session should implement, or a handover note.

In the entry header, click Start a new chat session anchored to this entry, or
right-click the entry in a phase and choose New Chat from this, or
choose AI › New Chat from Current Entry.

A new session opens with the entry path in the chat input. The AI reads the entry when you send the message.

Append file path to the chat input and Send to chat add the path to the current session instead.

Links in chat answers

The AI links entries with ce://projectdocs links. Clicking one opens the dashboard at that place.

Link	Opens
`ce://projectdocs?section=masterplan`	The masterplan.
`ce://projectdocs?section=phase&slug=<phase-slug>`	A phase.
`ce://projectdocs?section=entry&slug=<phase-slug>&path=<path>`	An entry.
`ce://projectdocs?section=decisions`	The decision records.

Tools

Read

Tool	Purpose
`projectdocs_list_phases`	All phases with status, order, parent and entry count.
`projectdocs_read_entry`	One entry, frontmatter and body.
`projectdocs_search_docs`	Full-text search with `#tag`, `§status` and `"phrase"`.
`projectdocs_semantic_search`	Search by meaning, also across languages. With `source: spec` or `all` it includes specs.
`projectdocs_for_symbol`	Entries anchored to a symbol.
`symbols_for_projectdoc`	Symbols anchored from an entry, with their current position.

Write (Brave mode)

Tool	Purpose
`projectdocs_create_phase`	New phase, optionally as a subphase.
`projectdocs_rename_phase`, `projectdocs_set_phase_parent`	Rename or move a phase.
`projectdocs_set_phase_status`, `projectdocs_mark_phase_reviewed`	Phase status, review time. The AI does not set `verified`.
`projectdocs_create_entry`	New entry of any type.
`projectdocs_append_entry_section`	Append a section, for example `Fix (date)`.
`projectdocs_set_entry_status`, `projectdocs_set_entry_tags`	Lifecycle status, tags.
`projectdocs_append_comment`, `projectdocs_toggle_comment`	Comment blocks.
`projectdocs_anchor_add`, `projectdocs_anchor_remove`	Symbol anchors. With `ref=<slug>` into a reference project.
`projectdocs_commit_anchor_add`, `projectdocs_commit_anchor_remove`	Commit anchors.
`projectdocs_sync_anchors`	Rebuild the local anchor index from the frontmatter of all entries.
`session_new_chat_from_entry`	Start a new chat session with an entry and a starting text.

Semantic search

projectdocs_semantic_search combines keyword ranking with embeddings computed on your machine. It finds an entry that answers a question in other words or in another language. The embedding model is downloaded once, about 130 MB, on first use.