MCP Tool Flow Map

"What kind of work am I doing?" → "Which tool gets it done?" 19 curated tools mapped by intent — capture, extract, search, research, annotate.

Every interaction with Multiplist falls into one of six categories. Pick the one that matches what you're trying to do — the tools inside it are already organized around that intent.

Tip: Example prompts are the best way to explore. Try "save this conversation to my vault," "what have I decided about pricing," or "research how my thinking on onboarding has evolved." Multiplist picks the right tool for each intent.

# 1. Capture — "I want to save something to my vault"

Everything starts here. You bring content in — a conversation, a document, a note — and Multiplist saves it as a Source. Extraction runs automatically unless you opt out.

Typical flow: create_source → (auto-extraction) → get_source

Intent	Tool	What it does
Save new content	`create_source`	Import a conversation, document, or note. Runs extraction automatically by default.
Update a saved source	`update_source`	Append or replace content. Only new material is re-extracted — efficient and additive.
Remove a source	`delete_source`	Remove a source and every seed extracted from it.

Example prompt: "Save this conversation to my vault." → create_source with autoExtract: true.

# 2. Extraction — "I want to extract meaning from a source"

Turn freeform content into structured, cited seeds. Multiplist extracts nine categories of meaning — decisions, frameworks, golden passages, definitions, exemplars, actions, questions, offers, and emergence — each seed traced back to exact character positions in the source.

Typical flow: trigger_extraction (single) · batch_extract → batch_status (many)

Intent	Tool	What it does
Extract one source	`trigger_extraction`	Run extraction on a single source. Use when you skipped auto-extraction or want to re-run with a specific skill.
Extract many sources	`batch_extract`	Kick off extraction across up to 50 sources at once. Returns a job ID.
Poll a batch job	`batch_status`	Check progress and completion state of a batch extraction job.

Example prompt: "Extract decisions and frameworks from that source I just saved." → trigger_extraction with categories: ["decision", "framework"].

# 3. Custom Skills — "I want to define what to look for"

Skills are custom extraction recipes. You define what matters in your domain, and Multiplist uses that definition during extraction. A coach creates "Client Breakthroughs." A developer creates "Architecture Signals." A researcher creates "Research Distillation."

Typical flow: create_skill → match_skills → trigger_extraction (with skillId)

Intent	Tool	What it does
Define a new skill	`create_skill`	Create a domain-specific extraction skill with its own schema and instructions.
Browse your skills	`list_skills`	List all extraction skills in your vault, optionally filtered by domain or container.
Read a skill	`get_skill`	Fetch a single skill's full instructions and schema.
Match skills to a source	`match_skills`	Find which skills apply to a given source based on content and trigger conditions.

Example prompt: "Create a skill that extracts client breakthroughs from my coaching session transcripts." → create_skill with schema + instructions.

# 4. Discovery — "I want to find something in my vault"

Search and retrieve across everything you've saved. Every result cites its source. Use search_vault to query meaning across all sources; search_sources to browse by metadata; get_source to pull a specific source back with its full Seed Doc.

Typical flow: get_vault_summary → search_vault | search_sources → get_source

Intent	Tool	What it does
Pull a specific source	`get_source`	Retrieve a source with its abstract, seeds grouped by category, and assets.
Search by meaning	`search_vault`	Search seeds matching a query — "what have I decided about pricing?" Returns cited results.
Browse the catalog	`search_sources`	Browse sources by date, type, status, or keyword. The card catalog.
Orient yourself	`get_vault_summary`	High-level snapshot: source count, category distribution, recent activity, topic clusters.
Find what's missing	`get_vault_gaps`	Sparse categories, unextracted sources, under-represented areas.

Example prompt: "What have I decided about onboarding this quarter?" → search_vault with category: "decision" and a temperature filter for recency.

# 5. Research — "I want to synthesize across my vault"

For questions that span many sources — how has your pricing strategy evolved, what are the emerging patterns across coaching sessions, where do your recent decisions contradict older ones — request a research brief. The brief synthesizes across your vault with citations to every source used. Runs asynchronously.

Typical flow: request_brief → get_brief_status (poll) → the brief is saved as its own source

Intent	Tool	What it does
Request a brief	`request_brief`	Synthesize across vault sources on a topic. Choose depth (quick, standard, deep) and brief type (research, evolution, constellation, status).
Poll a brief	`get_brief_status`	Check progress and completion. When done, the brief is saved as its own source in your vault.

Example prompt: "Research how my thinking about AI memory has evolved over the past three months." → request_brief with briefType: "evolution".

# 6. Annotate — "I want to write something back to a source"

Write corrections, evolution notes, and commentary back to a source. Annotations are the author's own voice — notes on their own work. Marginalia is later-mind marks — corrections, tensions, evolutions, or commentary added after the fact, possibly by a different scribe. Every mark records who made it and what kind of mark it is.

Typical flow: get_source → add_annotation (author's own) · add_marginalia (later mind)

Intent	Tool	What it does
Author annotation	`add_annotation`	Attach a real-time note by the source's own author while working on it. Records intent and confidence.
Later-mind mark	`add_marginalia`	Add a correction, tension flag, evolution narrative, or commentary. Records the scribe identity and mark type for cross-AI provenance.

When to use which: Use add_annotation when the source's author is marking their own work in the moment. Use add_marginalia when anyone (AI or human) is marking a source from the outside — fixing an earlier call, flagging a contradiction across sources, or narrating how a concept has shifted.

# All 19 tools at a glance

CAPTURE          create_source · update_source · delete_source
EXTRACTION       trigger_extraction · batch_extract · batch_status
CUSTOM SKILLS    create_skill · list_skills · get_skill · match_skills
DISCOVERY        get_source · search_vault · search_sources ·
                 get_vault_summary · get_vault_gaps
RESEARCH         request_brief · get_brief_status
ANNOTATE         add_annotation · add_marginalia

Every tool returns cited results. Every claim in Multiplist traces back to a specific source and character range — that's what makes it an IP Vault with Citations rather than a notes app.