M3 Memory MCP Server
m3-memory
PyPI
v2026.8.30.1
Published by skynetcmd — no publish provenance, so origin is unverified, but the source is public: the repository link below is self-declared yet readable, so you can inspect the code before adopting it.
Give your AI agent a memory — local-first, private, easy to install. Guided setup wizard; works out of the box on your own machine, no cloud. For everyday agent users, homelabs, and developers alike · 99.2% LongMemEval-S retrieval @ k=10 · Works with Claude · Gemini · Antigravity · OpenCode · OpenClaw · Hermes · any MCP agent (native + one-command plugins) · Hybrid search (FTS5 + vector + MMR) · GDPR · FIPS 140-3 deployment-ready · 100% local (fully offline) or cloud capable
The grade answers one question — how safe is this server for you to adopt — so it is computed in two auditable stages. Nothing below is an opinion or an LLM's guess; every line is a real term the deterministic engine applied, and the same input always yields the same number.
1. Threat score — 100 − 0 = 100. What the published surface and source actually contain:
The deterministic scan raised no scored threat in the surface it inspected — the threat score stayed at 100. Capability observations and advisory notes are recorded but never lower it.
2. Client adoption risk — 100 − 7 = 93. Three small, subtract-only factors that reflect your risk in adopting it — a clean scan proves less on a powerful, unverified or barely-inspectable package, so the grade says so plainly:
| Points | Adoption-risk factor |
|---|---|
| −6 | capability blast radius (high) — client exposure if the model is manipulated |
| −1 | publisher verification (public source) — no provenance, but the source is public and inspectable |
Capability observations and info notes are shown under Findings but never scored.
Open any row's finding below for the file, line and evidence behind a deduction.
Findings 14
high Shell/command execution in server code (bin/_task_runtime.py)MTC-SRC-002
In the server's implementation (`bin/_task_runtime.py:48`): Spawning a shell/process is command-execution capability; with unsanitized tool input it is command injection / RCE. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.
Evidence: rypoint makes: subprocess.run([...], **no_window_kwargs()) On POSIX this is an empty dict (CREATE_NO_WINDO
Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.
Location: server bin/_task_runtime.py
high Shell/command execution in server code (bin/aider_patch.py)MTC-SRC-002
In the server's implementation (`bin/aider_patch.py:31`): Spawning a shell/process is command-execution capability; with unsanitized tool input it is command injection / RCE. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.
Evidence: try: branch = subprocess.check_output( ['git', 'branch', '--show-current'], stderr=sub
Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.
Location: server bin/aider_patch.py
high Shell/command execution in server code (bin/auth_utils.py)MTC-SRC-002
In the server's implementation (`bin/auth_utils.py:118`): Spawning a shell/process is command-execution capability; with unsanitized tool input it is command injection / RCE. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.
Evidence: y: result = subprocess.run( ["security", "find-generic-password", "-s", "AGENT_OS_MASTER_KEY
Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.
Location: server bin/auth_utils.py
high Shell/command execution in server code (bin/chatlog_init.py)MTC-SRC-002
In the server's implementation (`bin/chatlog_init.py:238`): Spawning a shell/process is command-execution capability; with unsanitized tool input it is command injection / RCE. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.
Evidence: try: result = subprocess.run( [sys.executable, migrate_script, "up", "--target", "chatlog", "-y"],
Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.
Location: server bin/chatlog_init.py
high Shell/command execution in server code (bin/chatlog_status.py)MTC-SRC-002
In the server's implementation (`bin/chatlog_status.py:769`): Spawning a shell/process is command-execution capability; with unsanitized tool input it is command injection / RCE. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.
Evidence: try: result = subprocess.run(cmd, env=env, shell=False) print("=" * 80) print(f"Command finish
Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.
Location: server bin/chatlog_status.py
high Shell/command execution in server code (bin/check_control_chars.py)MTC-SRC-002
In the server's implementation (`bin/check_control_chars.py:68`): Spawning a shell/process is command-execution capability; with unsanitized tool input it is command injection / RCE. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.
Evidence: try: out = subprocess.run( ["git", "diff", "--cached", "--name-only", "--diff-filter=ACMR"],
Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.
Location: server bin/check_control_chars.py
high Shell/command execution in server code (bin/check_tool_catalog_drift.py)MTC-SRC-002
In the server's implementation (`bin/check_tool_catalog_drift.py:66`): Spawning a shell/process is command-execution capability; with unsanitized tool input it is command injection / RCE. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.
Evidence: tr]) -> int: return subprocess.run(cmd, cwd=_ROOT).returncode def _regenerate() -> bool: """Run all generators
Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.
Location: server bin/check_tool_catalog_drift.py
high Shell/command execution in server code (bin/dashboard_server.py)MTC-SRC-002
In the server's implementation (`bin/dashboard_server.py:2591`): Spawning a shell/process is command-execution capability; with unsanitized tool input it is command injection / RCE. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.
Evidence: w_kwargs proc = subprocess.Popen( cmd, env=env, stdout=log_file_handle,
Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.
Location: server bin/dashboard_server.py
high Shell/command execution in server code (bin/deep_sync.py)MTC-SRC-002
In the server's implementation (`bin/deep_sync.py:29`): Spawning a shell/process is command-execution capability; with unsanitized tool input it is command injection / RCE. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.
Evidence: up_script): subprocess.run([cleanup_script], check=True, stdout=subprocess.DEVNULL) logger.info(
Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.
Location: server bin/deep_sync.py
high Shell/command execution in server code (bin/doctor/claude_mcp_probe.py)MTC-SRC-002
In the server's implementation (`bin/doctor/claude_mcp_probe.py:168`): Spawning a shell/process is command-execution capability; with unsanitized tool input it is command injection / RCE. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.
Evidence: try: proc = subprocess.run( ["claude", "mcp", "remove", "--scope", "user", name], ch
Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.
Location: server bin/doctor/claude_mcp_probe.py
medium Dynamic module load from a non-literal (bin/mem0_scan.py)MTC-SRC-005
In the server's implementation (`bin/mem0_scan.py:270`): Loading a module chosen at runtime (from a variable) can pull in and run attacker-influenced code paths. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.
Evidence: f"{totals['import']} import(s) · {totals['drop-in']} drop-in call(s) · " f"{totals['mapped']} mapped · {total
Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.
Location: server bin/mem0_scan.py
medium Dynamic module load from a non-literal (m3_memory/cli.py)MTC-SRC-005
In the server's implementation (`m3_memory/cli.py:601`): Loading a module chosen at runtime (from a variable) can pull in and run attacker-influenced code paths. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.
Evidence: try: __import__(mod) except ModuleNotFoundError: missing.append(mod)
Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.
Location: server m3_memory/cli.py
medium Dynamic module load from a non-literal (m3_memory/installer.py)MTC-SRC-005
In the server's implementation (`m3_memory/installer.py:754`): Loading a module chosen at runtime (from a variable) can pull in and run attacker-influenced code paths. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.
Evidence: try: __import__(mod) except ModuleNotFoundError: return False return True d
Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.
Location: server m3_memory/installer.py
medium Dynamic module load from a non-literal (m3_memory/setup_wizard.py)MTC-SRC-005
In the server's implementation (`m3_memory/setup_wizard.py:2488`): Loading a module chosen at runtime (from a variable) can pull in and run attacker-influenced code paths. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.
Evidence: try: __import__(mod) except ModuleNotFoundError: have = False break
Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.
Location: server m3_memory/setup_wizard.py
Tools 46
Each tool and what it can reach — statically extracted from the published source.
files_corpus_getingests untrusted inputfiles_getingests untrusted inputget_news_headlinesingests untrusted inputm3_web_searchingests untrusted inputweb_searchingests untrusted inputcheck_thermal_loadno sensitive capabilitydebug_analyzeno sensitive capabilitydebug_bisectno sensitive capabilitydebug_correlateno sensitive capabilitydebug_historyno sensitive capability
Show 36 more tools ↓
debug_reportno sensitive capabilitydebug_traceno sensitive capabilityfiles_corpus_createno sensitive capabilityfiles_corpus_deleteno sensitive capabilityfiles_corpus_listno sensitive capabilityfiles_corpus_setno sensitive capabilityfiles_dedupno sensitive capabilityfiles_dedup_listno sensitive capabilityfiles_dedup_reviewno sensitive capabilityfiles_entity_coalesceno sensitive capabilityfiles_entity_coalesce_applyno sensitive capabilityfiles_entity_coalesce_listno sensitive capabilityfiles_entity_coalesce_reviewno sensitive capabilityfiles_entity_coalesce_unapplyno sensitive capabilityfiles_extract_pendingno sensitive capabilityfiles_healthno sensitive capabilityfiles_indexno sensitive capabilityfiles_ingestno sensitive capabilityfiles_link_renameno sensitive capabilityfiles_promotableno sensitive capabilityfiles_promoteno sensitive capabilityfiles_promotion_listno sensitive capabilityfiles_searchno sensitive capabilityfiles_staleness_reviewno sensitive capabilityfiles_statsno sensitive capabilityfiles_watch_onceno sensitive capabilityforgetno sensitive capabilitygrok_askno sensitive capabilitylog_activityno sensitive capabilityquery_decisionsno sensitive capabilityquery_local_modelno sensitive capabilityrecallno sensitive capabilityrememberno sensitive capabilityretire_focusno sensitive capabilitysave_handoffno sensitive capabilityupdate_focusno sensitive capability
What this scan could not see
- Tools were statically extracted from the published source (46 recovered), not enumerated from a running server. Tool-poisoning, Unicode-smuggling, capability and toxic-flow analysis ran on this inferred surface, but a mis-parsed registration could be missed or mis-attributed, so tool-derived findings are capped below “confirmed”. To grade the real runtime surface, scan the running server: --command "npx -y <package>".
Versions 10
Scan history per published version. The engine is deterministic — the same version always yields the same score, so a changed score means the package itself changed.
| Version | Score | Findings | Engine | Scanned |
|---|---|---|---|---|
v2026.8.30.1 latest |
A 93/100 | 14 | 1.13.0 | 2026-09-01 |
v2026.8.30.0 |
A 93/100 | 14 | 1.13.0 | 2026-08-31 |
v2026.8.29.0 |
A 93/100 | 14 | 1.13.0 | 2026-08-30 |
v2026.8.19.16 |
A 93/100 | 14 | 1.13.0 | 2026-08-25 |
v2026.8.19.15 |
A 93/100 | 14 | 1.12.1 | 2026-08-12 |
Show 5 more versions ↓
v2026.8.19.10 |
A 93/100 | 14 | 1.12.1 | 2026-08-10 |
v2026.8.19.2 |
A 93/100 | 14 | 1.12.1 | 2026-08-09 |
v2026.7.31.1 |
A 93/100 | 14 | 1.12.1 | 2026-08-01 |
v2026.7.30.1 |
A 93/100 | 14 | 1.12.1 | 2026-07-31 |
v2026.7.28.1 |
A 93/100 | 14 | 1.12.1 | 2026-07-29 |
Embed this score
Show this server's live Trust Score in your README, docs or website. The badge is served straight from the registry and updates automatically after every rescan — no API key needed. It links back to this page, so anyone who sees the grade can also read the findings behind it instead of taking a number on faith.
Verify this score yourself
The score above is reproducible: the same package version always yields the same result. Run it locally or over the free API — no account, no LLM, fully deterministic.
npx mcptrustchecker scan m3-memory --online --registry pypiMore in Productivity & Workflow
MCP AgentChat - 让 AI Agent 通过 Telegram 与用户实时交互：发送消息、图片，等待用户回复
A Model Context Protocol (MCP) server for Airthings Air Quality Monitor devices.
MCP server for Anki
TypeScript package for reading and searching Apple Notes on macOS via direct SQLite access. Includes markdown conversion, attachment support, and offers a local MCP server!
Verified biotech catalyst calendar (PDUFA/AdComm/trial readouts) anchored to official sources.
Appointment booking platform letting agents check provider availability and create bookings.