Atlas / MCP servers / plateerlab / Synaptic Memory

Synaptic MemoryCAUTION

Name: Synaptic Memory
Author: plateerlab

mcp/plateerlab/synaptic-memory

MCP server by plateerlab · agent memory · ai agent · change data capture · Audited 2026-10-08

⚑ Report / re-scan Source ↗

Knowledge graph + MCP tool server for LLM agents with hybrid retrieval, live DB sync, Korean FTS, and memory feedback.

Verdict

CAUTION

Grade

Trust score

61 /100

Exposed tools

43 29r · 11w · 3d

Transport

—

License

NOASSERTION

Stars

Overview

From the repository's own README, as read at the audited commit. Badges and raw HTML are left out.

Default path: zero API calls at index time. Zero infra. Zero lock-in. A knowledge graph + MCP tool server for LLM agents, with hybrid retrieval, CDC-based live database sync, and Korean FTS built in.

[](https://pypi.org/project/synaptic-memory/) [](https://pypi.org/project/synaptic-memory/) [](LICENSE)

한국어 README

5-minute start

pip install "synaptic-memory[sqlite,korean,vector]"
synaptic-quickstart --db quickstart.db

That command builds a tiny SQLite-backed graph and runs three searches — all without calling any LLM at indexing time. Omit --db for an in-memory, zero-dependency smoke test. Full source for the expanded example: examples/quickstart.py.

Why not just RAG?

Plain RAG usually answers from independent chunks. Synaptic builds a graph first, so an agent can search, follow relations, inspect structured rows, and remember which evidence helped.

It is not a vector database replacement. It is the graph and tool layer around your existing documents, SQL data, embedding endpoint, and agent runtime.

Build and search

import asyncio
from synaptic import SynapticGraph

async def main():
# Any data → knowledge graph (CSV, JSONL, directory)
graph = await SynapticGraph.from_data("./my_data/", preset="rag")
try:
result = await graph.search("my question")
print(result.nodes[0].

Read from source at commit a42a99ee2960OBSERVED · 2026-10-08

Connect

Built from this server's own package name, version and transport as found in its source — not copied from anyone's documentation, so it cannot drift against a page we do not control. Replace the environment placeholders with a token scoped to the least it needs.

claude-code

claude mcp add synaptic-memory --env ANTHROPIC_API_KEY=${ANTHROPIC_API_KEY} --env DEEPSEEK_API_KEY=${DEEPSEEK_API_KEY} --env OPENAI_API_KEY=${OPENAI_API_KEY} -- uvx synaptic-memory

claude-desktop

{
  "mcpServers": {
    "synaptic-memory": {
      "command": "uvx",
      "args": [
        "synaptic-memory"
      ],
      "env": {
        "ANTHROPIC_API_KEY": "${ANTHROPIC_API_KEY}",
        "DEEPSEEK_API_KEY": "${DEEPSEEK_API_KEY}",
        "OPENAI_API_KEY": "${OPENAI_API_KEY}"
      }
    }
  }
}

Exposed tools (43)

29 read · 11 write · 3 destructive. Blast radius: 3 tools can delete or overwrite — an agent that can be talked into calling a tool can be talked into calling this one.

Tool	Risk	Description
`agent_aggregate_nodes`	read	Aggregate nodes by property — GROUP BY + COUNT/SUM/AVG.
`agent_compare_search`	read	Compare search — decompose multi-topic query and search in parallel.
`agent_count`	read	Count how many nodes match a filter without fetching them.
`agent_deep_search`	read	Deep search — search + expand + read documents in ONE call.
`agent_expand`	read	Walk one graph hop out from a specific node.
`agent_explore_context`	read	Explore the knowledge graph around a specific node, following semantic relationships.
`agent_filter_nodes`	read	Filter nodes by property value — for structured/tabular data.
`agent_find_similar`	read	Search knowledge with agent-aware intent for smarter results.
`agent_follow`	read	Walk a specific edge type from a starting node.
`agent_get_document`	read	Fetch a document with smart context control.
`agent_get_reasoning_chain`	read	Get the full reasoning chain for a decision: decision → outcome → lessons learned.
`agent_join_related`	read	Follow a foreign key to find related records.
`agent_list_categories`	read	List all top-level categories in the knowledge graph.
`agent_log_action`	read	Log a tool call or action within an agent session.
`agent_record_decision`	read	Record a decision made by the agent with rationale and considered alternatives.
`agent_record_outcome`	read	Record the outcome of a previous decision. Triggers Hebbian learning.
`agent_search`	read	Multi-turn search — finds evidence for a natural-language query.
`agent_search_exact`	read	Literal substring match for codes, IDs, or exact strings.
`agent_session_info`	read	Inspect the current state of an agent session.
`agent_start_session`	write	Start an agent work session. All subsequent actions can be linked to this session.
`agent_top_nodes`	read	Return the top-N rows of ``table`` ordered by ``sort_by``.
`knowledge_add`	write	Add a new knowledge node to the graph.
`knowledge_add_chunks`	write	Ingest pre-chunked content (BYO-chunker workflow).
`knowledge_add_document`	write	Add a long document to the graph with automatic chunking.
`knowledge_add_table`	write	Ingest a structured table into the graph.
`knowledge_ask`	read	Answer a question end-to-end with honest routing — cheap path first, agent only when needed.
`knowledge_backfill`	read	Repair existing nodes that are missing embeddings or phrase hubs.
`knowledge_consolidate`	write	Run memory consolidation — expire old L0 nodes, promote accessed ones.
`knowledge_export`	read	Export the knowledge graph.
`knowledge_ingest_path`	read	Ingest a file from the local filesystem into the current graph.
`knowledge_link`	write	Create a link between two knowledge nodes.
`knowledge_merge_nodes`	destructive	Merge ``drop_id`` into ``keep_id`` and delete ``drop_id``.
`knowledge_reinforce`	read	Reinforce knowledge nodes after use (Hebbian learning).
`knowledge_remove`	destructive	Delete a single node and cascade-remove its edges.
`knowledge_search`	read	Search the knowledge graph for lessons, decisions, patterns, and past outcomes.
`knowledge_snapshot`	read	Generate a markdown snapshot of the graph — for agent priming.
`knowledge_stats`	read	Get knowledge graph statistics — node counts by kind and level, cache stats.
`knowledge_sync_from_database`	write	Incrementally sync the graph with a live database (CDC).
`knowledge_unlink`	destructive	Delete edges from ``source_id`` to ``target_id``.
`knowledge_update`	write	Update fields of an existing knowledge node.
`knowledge_update_edge`	write	Change weight and/or kind of existing edge(s) between two nodes.
`ontology_define_type`	write	Define or update a custom node/edge type in the ontology.
`ontology_query_schema`	read	Query the ontology schema. Returns type definitions including inherited properties.

Trust audit

CAUTIONgrade D · trust 61/100 Install with care. The audit found things worth knowing before you trust its output.

Layer	What it checks	Result
L0	Provenance & inventory	WARN
L1	Static analysis of the code	WARN
L2	Instruction surface (what it tells the agent)	PASS
L3	Class-specific surface	WARN
L4	Behavioural (sandbox)	SKIPPED

What the source does

Filesystem: declared (5 observation(s))
Network: declared (6 observation(s))
Shell: declared (1 observation(s))
Dependencies: pinned
Secrets in source: found

Findings (25)

MEDIUMInventory / provenance · inv.binary · CWE-1104

eval/data/gt_datasets.xlsx

gt_datasets.xlsx

Why it matters. a compiled or binary member cannot be reviewed from source