A visual context graph for AI agents.

Your AI history, documents, artifacts, memory, rules, plans, research, summaries, and EoCs, structured for humans and agents. Context Vault turns the work you already did with AI into a one-way local mirror and Second Brain with an interactive graph layer other agents can navigate.

One-way local mirror. Source-grounded Markdown. High-signal indexes. Visual context graph. No local edits imported back into GRASPPY.

context-vault/ Agent-ready
CONTEXT.mdThe root routing map agents open first.
manifest
indexes/high-signal-turns.tsvImportance scores, keywords, entities, topics, and source paths.
fast search
graph/graph.htmlInteractive node map with communities, search, filters, node details, and source tracing.
visual
conversations/ summaries/ eoc/Full threads, turn summaries, and latest closure reports.
memory
artifacts/ documents/ analysis/Generated work, imported sources, and cleaned pipeline outputs.
sources

AI work disappears into tool silos and flat exports.

Modern AI work is not one chat. It is a trail of decisions, prompts, code, documents, research, generated artifacts, corrections, summaries, and handoffs spread across multiple assistants.

Context gets trapped

Important decisions stay inside individual AI tools where other systems cannot reliably find them.

Raw exports are not memory

A folder full of undifferentiated chats is technically portable, but agents still have to guess what matters and how ideas connect.

Agents need routing and relationships

Good agents do not need everything at once. They need a map, indexes, source paths, importance scores, and a graph of what connects to what.

Not a database dump. An intelligent routing and visual graph layer.

GRASPPY reads the structured knowledge it already maintains, renders it as a one-way local mirror, builds graph relationships from source metadata, and gives you a visual map written only inside your local grasppy-exports/context-vault/ folder.

CONTEXT.md
purpose: agent routing map
start_here: indexes/high-signal-turns.tsv
source_paths:
  conversations/
  summaries/
  artifacts/
  documents/
  analysis/
  eoc/
  graph/
safety: one_way_mirror
1
GRASPPY captures and analyzes AI work.Conversations, imported documents, artifacts, decisions, summaries, entities, topics, plans, memory, and research remain structured in GRASPPY.
2
Context Vault renders a local mirror.Files are written incrementally, using manifests and checksums so sync does not rewrite the whole vault every time.
3
Agents and humans use the graph before reading full sources.Importance scores, keywords, entity indexes, topic indexes, decision indexes, communities, and visual relationships help narrow the search fast.

Context Vault shows the map before an agent asks.

The graph layer maps conversations, high-signal turns, summaries, artifacts, documents, EoCs, entities, topics, keywords, decisions, and source paths into a searchable visual relationship map.

Interactive graph view

Zoom, pan, search, filter communities, inspect nodes, and trace how conversations, artifacts, documents, topics, and decisions connect.

Graph data for agents

Under the visual map, graph nodes and edges stay available as structured files agents can query before opening full source Markdown.

Confidence and signal

Direct database relationships are marked as extracted, weaker relationships are marked as inferred or ambiguous, and importance scores keep retrieval focused.

The parts of your AI work that usually get lost.

Conversations

One Markdown file per conversation, with source IDs, dates, platform, summaries, and optional full turns.

Artifacts and documents

Generated files and imported sources become organized Markdown with source-grounded metadata.

Pipeline analysis

Executive summaries, technical summaries, process flows, entities, topics, and decisions become cleaned outputs.

Turn intelligence

Turn summaries, importance scores, keywords, entity mentions, and topic signals are compiled into searchable index files.

EoC reports

Session closure reports are separated from ordinary turns so agents can quickly understand the final state of a work session.

Graph and manifests

CONTEXT.md, MANIFEST.md, graph files, schemas, and category manifests tell agents and graph tools where to go and what each folder contains.

Context Vault makes GRASPPY useful beyond GRASPPY.

The value is not just that your data is exportable. The value is that your AI work becomes structured enough for the next AI system to use it.

Capability
Raw chat export
Traditional notes
GRASPPY Context Vault
Search path
Scan everything.
Depends on manual organization.
Start with manifests, high-signal indexes, then the visual graph.
Signal quality
Mixed raw content.
Whatever the user remembered to write.
Summaries, importance scores, keywords, entities, topics, and source paths.
Agent portability
Possible, but inefficient.
Readable, but not built for routing.
Built as a local routing layer and visual context graph for AI agents.

GRASPPY remains the source of truth.

Context Vault ships with its own visual graph viewer, while keeping the local folder structure friendly to external graph tools. GRASPPY owns the source data, metadata, summaries, importance scores, and relationships.

Native first

The built-in graph uses GRASPPY’s structured database and indexes instead of re-discovering relationships from raw files.

Tool friendly

Graph files, manifests, and routing notes make the vault easy for graph tools and coding agents to inspect without special lock-in.

No hidden mutation

External tools can read the vault, but local graph exploration never changes GRASPPY records or imports edits back silently.

Ownership without turning local files into a risk.

Context Vault is designed as a one-way mirror. That is a deliberate product choice: local files should help agents read your context, not silently mutate your database.

One-way writes

GRASPPY renders Markdown from existing records and writes generated files locally. It does not import local edits back.

Generated-file registry

Cleanup is based on a registry of generated files, so “remove all” removes the mirror without deleting GRASPPY data.

User-controlled depth

Users decide whether to include full conversation turns, imported document bodies, and artifact content.

Build your AI Second Brain on structured context.

Capture the work. Preserve the decisions. Give your agents a map they can actually follow.