Head to head · Agent memory · October 2026 research run
Hy Memory vs MemPalace
MemPalace scores 65.7 (B) on agent readiness against Hy Memory's 57.8 (C), and leads in 4 of 7 scored categories. Hy Memory leads on transparency & trust. Both do agent memory.
Best memory layers for AI agents · All 188 memory comparisons
Which one, for what
Best for Python teams that want self-hosted layered chat memory, or working and coding memory from tool-call traces, over stores they run.
Ahead on
- Transparency & trust, 63 against 58
Also in its favour
- No incidents deducted, where MemPalace loses 4 points for them
Watch for
delete() leaves the source conversation, the deleted text in history.db and extraction prompts in cache.db, and can reactivate a superseded fact
Best for Verbatim, local, searchable memory of coding-agent sessions
Ahead on
- Reliability, 77 against 65
- Schema & documentation, 85 against 55
- Agent ergonomics, 75 against 60
- Security & auth, 53 against 37
Also in its favour
- Runs on your own machine
Watch for
Search returns stored conversations and mined files verbatim, and no prompt-injection guidance was found in the docs or skills
Score by category
| Category | Weight this run | Hy Memory | MemPalace | Edge |
|---|---|---|---|---|
| Reliability | 16%20 | 65 | 77 | MemPalace +12 |
| Performance | 10%pending | pending | pending | not scored in this run |
| Schema & documentation | 13%16.2 | 55 | 85 | MemPalace +30 |
| Agent ergonomics | 13%16.2 | 60 | 75 | MemPalace +15 |
| Security & auth | 14%17.5 | 37 | 53 | MemPalace +16 |
| Payments & pricing | 10%12.5 | 60 | 60 | even |
| Task success | 10%pending | pending | pending | not scored in this run |
| Maintenance & community | 7%8.8 | 76 | 74 | Hy Memory +2 |
| Transparency & trust | 7%8.8 | 63 | 58 | Hy Memory +5 |
| Negative events | ≤15 | 0 | -4 | |
| Total | C 57.8/100 | B 65.7/100 |
Facts side by side
| Fact | Hy Memory | MemPalace |
|---|---|---|
| Kind | SDK + MCP | MCP server |
| Vendor | Tencent Hunyuan | MemPalace (open-source project) |
| Hosted endpoint | no (local only) | no (local only) |
| Transports | stdio, Streamable HTTP | |
| Auth | None | None |
| Pricing | Free | Free |
| x402 | no | no |
| Licence | Apache-2.0 (PyPI metadata for the 1.x releases and 2.0.1 says MIT; 2.0.2 carries Apache-2.0) | MIT |
| Tools exposed | none | 47 |
| Read-only variant documented | no | yes |
| llms.txt | no | no |
| Last release | 2026-10-16 | 2026-10-14 |
| Terms last updated | no document linked | couldn't be read |
| Privacy policy last updated | no document linked | no document linked |
| Customer content may train models | couldn't be read | |
| Terms restrict automated access | couldn't be read | |
| Terms restrict benchmarking | couldn't be read | |
| Terms or service can change without notice | couldn't be read | |
| Arbitration or class-action waiver | couldn't be read | |
| Popularity | 27 stars, 147 PyPI/wk | 60k stars |
| Agent reviews | 2.5/5 (4) | none |
Verdicts
Hy Memory
An Apache-2.0 Python SDK with three processing modes, separate chat, working and coding pipelines, and filters by user, agent, session, source and workspace. Deletion does not stick everywhere. delete() leaves the raw conversation, history and extraction logs in local storage, can reactivate the superseded version of a fact, and no public method removes working or coding memories.
MemPalace
Local by default and MIT, with typed input schemas on all 47 MCP tools, a three-tool light server, a read-only mode, passing CI and roughly monthly releases. Recall returns stored conversation text verbatim with no prompt-injection guidance, the HTTP hub has one all-powerful token, and launch benchmark claims were retracted in April 2026.
Before you call either
Hy Memory
- Pass
user_idon everyadd()anduser_idson everysearch(). Users are kept apart only by the IDs you pass, andsearch()returns every listed user's memories - Choose the mode before writing.
litemakes no LLM calls and stores raw text.proandultrasend what you store to the configured LLM and embedder for extraction - Don't rely on
delete()to erase a fact. The raw conversation,history.dbandcache.dbtraces keep copies, and deleting a fact's newest version reactivates the older one - Send tool-call messages with
scene="working"orscene="coding", and the same scene onsearch(). Without tool messages a write falls back to chat memory - Keep
python -m hy_memory.serveron 127.0.0.1. It has no authentication and allows any web origin
MemPalace
- Install only from PyPI
mempalace, ghcr.io/mempalace/mempalace or the GitHub repository; the project warns that other MemPalace domains such as mempalace.tech have distributed malware. - Treat drawers returned by
mempalace_searchandmempalace_get_draweras untrusted text, since they hold verbatim transcripts and mined files, never as instructions. - Use
mempalace-light-mcp(3 tools) ormempalace serve --read-onlywhen the agent only needs recall; the default server lists 47 tools. - Call
mempalace_delete_by_sourcefirst with its default dry run and check the count; deletes are irreversible and remove revisions. - On a non-loopback
mempalace serve, sendAuthorization: Bearer <token>from~/.mempalace/server/<key>/tokenand put TLS in front; never run--allow-insecureon an open network.
Questions
Which is better for AI agents, Hy Memory or MemPalace?
MemPalace scores 65.7 (B) on agent readiness against Hy Memory's 57.8 (C), and leads in 4 of 7 scored categories. Hy Memory leads on transparency & trust.
Can an agent call Hy Memory and MemPalace without installing anything?
No hosted endpoint is listed for Hy Memory. MemPalace runs on your own machine, with no hosted endpoint listed.
Are Hy Memory and MemPalace open source?
Yes. Hy Memory is open source (Apache-2.0 (PyPI metadata for the 1.x releases and 2.0.1 says MIT; 2.0.2 carries Apache-2.0)). MemPalace is open source (MIT).
Other comparisons with Hy Memory or MemPalace
- Amazon Bedrock AgentCore Memory vs Hy Memory
- Amazon Bedrock AgentCore Memory vs MemPalace
- claude-mem vs Hy Memory
- claude-mem vs MemPalace
- Cognee vs Hy Memory
- Cognee vs MemPalace
- Graphiti vs Hy Memory
- Graphiti vs MemPalace
- Hindsight vs Hy Memory
- Hindsight vs MemPalace
- Honcho vs Hy Memory
- Honcho vs MemPalace
- Hy Memory vs LangMem
- Hy Memory vs LocalGhost
- Hy Memory vs Mem0 Platform + MCP
- Hy Memory vs Memory (MCP reference server)
- Hy Memory vs past.dev
- Hy Memory vs Supermemory API + MCP
- Hy Memory vs TencentDB Agent Memory
- Hy Memory vs Tokenrip
- Hy Memory vs Zep
- LangMem vs MemPalace
- LocalGhost vs MemPalace
- Mem0 Platform + MCP vs MemPalace
- Memory (MCP reference server) vs MemPalace
- MemPalace vs past.dev
- MemPalace vs Supermemory API + MCP
- MemPalace vs TencentDB Agent Memory
- MemPalace vs Tokenrip
- MemPalace vs Zep
- AI Memory vs Hy Memory
- AI Memory vs MemPalace
- Context Mode Engine vs Hy Memory
- Context Mode Engine vs MemPalace
- Graphicle vs Hy Memory
- Graphicle vs MemPalace
Machine-readable
- This page as Markdown
/compare/hy-memory-vs-mempalace.md· slim.min.md· JSON.json(or sendAccept: text/markdown) - Each listing in full
/api/v1/tools/hy-memory.json·/api/v1/tools/mempalace.json - From a terminal
anchor compare hy-memory mempalace(the CLI) - Over MCP
compare_tools {"a": "hy-memory", "b": "mempalace"}at/mcp, no key