Omni-Context exposes a local knowledge and decision store to assistants through a stdio bridge or authenticated loopback HTTP endpoint.
Omni-Context exposes a local knowledge and decision store to assistants through a stdio bridge or authenticated loopback HTTP endpoint.
The project combines a desktop application and a brain server for captures, knowledge retrieval and decision context. Its canonical tool manifest defines the actual MCP surface. Use that manifest rather than an older tool count or names copied from another memory server.
The desktop installation uses mcp-proxy.js, which reads the local token and forwards requests to the running app. A standalone source server uses mcp-server.js with DB_PATH selecting its SQLite store. The documented HTTP route is a local /mcp endpoint with a bearer token; localhost does not remove its authentication requirement.
Save a small test conclusion and retrieve it from the next session. Confirm that the desktop app and client are using the intended store before diagnosing a missing memory. A stale decision should be revised with its supporting context, not treated as a permanent instruction. Keep capture and write operations scoped to information the user intends to retain.
Check the linked publisher documentation and your connected client's tool list. Names and availability can vary with the server version and account permissions.
| Tool | What it does |
|---|---|
| record_capture | Store a capture event (screenshot, clipboard content, system logs) into the knowledge graph. Use when the user explicitly wants to save a moment or piece of context for future reference. |
| get_core_context | Get a compact set of the user's core principles. Pass the current topic as query when possible so only relevant principles are returned; without a query the server returns a capped general overview, not the full core set. |
| search_entities | Search for entities in the knowledge graph by name or description. Use when you need to find specific concepts, tools, people, projects, or code the user has encountered before. |
| add_entity | Add a new entity to the knowledge graph. Use when you discover a concept, tool, person, pattern, or piece of information worth remembering for future conversations. |
| get_entity | Retrieve full details and all relationships for a specific entity by its ID. Use when you have an entity ID from search results and need to explore its connections. |
| add_relationship | Create a relationship edge between two entities in the knowledge graph (e.g. depends_on, conflicts_with, extends, belongs_to). Use when you discover a meaningful connection between two concepts. |
| get_graph_neighborhood | Get the graph neighborhood around a specific entity—its directly related entities and their connections. Use when you need to understand the context and ecosystem surrounding a concept, rather than just its properties. |
| extract_from_capture | Analyze raw text or captured content and automatically extract entities, relationships, and principles into the knowledge graph. Use when processing a saved capture or analyzing a block of user-provided content. |
| list_entities | List all entities in the knowledge graph, optionally filtered by type. Use to get an overview of what the user has stored, or to browse entities of a specific category. |
| update_entity | Update an existing entity's name, description, tags, or metadata. Use when you need to correct or enrich information about something already in the knowledge graph. |
| get_stats | Get system statistics: total entity count, relationship count, entity type distribution, and memory usage. Use when you need a high-level overview of the knowledge graph's size and composition. |
| vector_search | Search the knowledge graph by semantic similarity using vector embeddings. Unlike text search, this finds conceptually related entities even when they use different words. Use when a keyword search misses relevant results. |
| unified_memory_search | Fused memory search combining text search, vector similarity, and graph traversal. Returns the most relevant entities for a given query across all three methods, deduplicated. This is the recommended go-to search when you want the most complete results. |
| get_decision_context | When you are facing a specific situation or decision and need context from the user's knowledge graph to inform your judgment. Returns relevant principles, related memories, historical conflicts, and graph neighborhood — but does NOT make the decision for you. Call this before giving advice that depends on the user's past patterns or preferences. |
| save_conclusion | Call this when the conversation reaches a meaningful conclusion, a decision is made, or the user learns something worth remembering. ALWAYS proactively call this at the end of a substantive conversation to save key takeaways into long-term memory. Pass a short summary and (if relevant) related entity IDs from earlier search results. |
| save_decision | Save a decision the user just made into the knowledge graph. Creates a decision entity linked to cited principles/memories via decision_referenced edges, and archives the full context for future retrieval. Use when the user has made a concrete choice and wants it remembered. |
| analyze_decision | Analyze a decision situation using the knowledge graph as context. Retrieves relevant principles, history, and conflicts, then feeds them to an LLM for structured analysis including pros/cons, risks, recommendation, and evidence citations. Use when the user wants AI-assisted decision analysis, not just raw data. |
| discuss_decision | Continue a decision discussion with AI. Takes the conversation history and returns a follow-up response grounded in the knowledge graph context. Use for multi-turn decision refinement. |
| get_decision_lineage | Trace the lineage of a decision — what sources it cited and what other decisions form a chain with it. Useful for understanding how past decisions connect to form a reasoning path. |
| record_decision_outcome | Record an observed result for a saved decision, including calibration, failed assumptions, lessons, and follow-up actions. This never changes core principles automatically. |
| get_decay_report | Get the memory decay report showing which entities have decayed over time and may need reinforcement or cleanup. Use when maintaining the knowledge graph's health. |
| ask_memory | Ask the user's second brain a natural-language question. Retrieves relevant memories from the local knowledge graph and answers grounded in them, citing sources. Use for "what do I know/think about X" style questions. |
| graph_answer | Answer a question or help with a decision using the user's knowledge graph. Returns a structured result: a one-line conclusion, supporting reasons (each citing memory nodes), the cited sub-graph edges, and—for decisions—clarifying questions. Prefer for decisions ("should I…", "which…") and whenever grounded, citable reasoning is wanted. |
| merge_entities | Soft-merge two duplicate entities: keep one, fold the other into it (relationships repointed, the other hidden). Reversible. Use to consolidate duplicates instead of deleting. |
| set_core_principle | Mark a principle as core (injected into every AI conversation) or demote it from core. Keep core to the user few distinctive operating principles; demote generic or bulk-imported ones. |
| delete_entity | Permanently delete an entity and its relationships. Irreversible — prefer merge_entities for duplicates. Use only for clearing genuine junk/wrong memories. |
{
"mcpServers": {
"omni-context": {
"command": "node",
"args": [
"/ABSOLUTE_PATH/Omni-context/brain-server/dist/mcp-proxy.js"
]
}
}
}The running Omni-Context desktop app and built proxy, or a configured standalone brain server and SQLite path.
Store entities, relations and observations in a local knowledge graph with the reference Memory MCP server.
A structured scratchpad for hard problems — stepwise reasoning with revisions, branches and visible logic.
Symbol-level code navigation, refactoring and memory for coding agents — the IDE brain your assistant has been missing.
Stop letting your assistant hallucinate n8n node parameters — this server hands it the real schemas, templates and validation.
Chat with your second brain — search, read and write vault notes through the Local REST API.
Enables AI assistants to securely interact with Azure DevOps through a standardized protocol.