Help your Agent remember fixed workflows, accumulate past experience, and reuse user preferences — **so people can pull their attention out of repetitive work and back to creation, judgment, and life itself**.
**Plugged into OpenClaw**, it saves up to **61.38% tokens**, lifts pass rate by **+51.52%** (relative), and pushes PersonaMem accuracy from **48%** to **76%**.
> These are long-session evaluations, not single-turn isolated runs. Multiple tasks are concatenated into the same session and executed back-to-back. For example, each SWE-bench session runs 50 tasks consecutively to simulate the context-accumulation pressure faced by a real long-horizon Agent.
TencentDB Agent Memory helps the Agent learn your workflows, retain task context, and reuse past experience, so that humans can spend their attention on judgment, creation, and work that actually matters.
In real work, plenty of things should not need to be re-explained: fixed SOPs, common analysis conventions, project background, user preferences, and so on. We capture these stable experiences and let the Agent reuse them automatically when the moment is right.
TencentDB Agent Memory is not a plain chat log, not a one-way irreversible summary, and not just RAG. We design memory as a layered information management system: the database carries the searchable, filterable, recallable factual base; the file system carries the readable, editable, progressively-disclosable task canvases, scene blocks, and user personas. Short-term compression solves in-task information overload, while long-term personalized memory solves cross-session user understanding.
In long tasks, what eats your context window is rarely the user's goal — it is the byproducts of tool calls: search results, web page bodies, file chunks, test logs, error traces, diffs, intermediate versions. TencentDB Agent Memory offloads these full payloads to external files, keeping only summaries, paths, and task state near the active context.
The point here is not "deleting history" but "folding history": most of the time the Agent looks at the task map; when it needs detail it drills down through `node_id` and `result_ref` back to the raw evidence.
Cross-session memory is a different problem. Raw conversation logs are a low-density ore: they contain preferences, facts, emotions, long-term goals — and a lot of noise. Plain vector retrieval can only find "similar fragments" and rarely surfaces the user's stable, long-term traits.
The upper layers help the Agent "understand you"; the lower layers back it up with factual detail. The Agent gets both the high-level read and the receipts when needed.
Once enabled, TencentDB Agent Memory automatically handles conversation capture, memory extraction, scene aggregation, persona generation, and recall before the next turn.
- **Short-term compression** solves "the current task is too long": offload raw tool results to external storage, fold the task structure into a Mermaid canvas.
- **Long-term personalized memory** solves "next time we meet, you don't know me": refine raw conversations layer by layer into structured memories, scene blocks, and a user persona.
Both share the same engineering principle: **lower layers preserve evidence, upper layers preserve structure; you read upper layers by default and drill down only when needed.**
| Short-term compression | `refs/*.md` raw tool result | `offload-*.jsonl` tool summary | `mmds/*.mmd` Mermaid task canvas / metadata | Keep long tasks moving instead of being dragged down by context |
| Long-term personalized memory | L0 raw conversation | L1 structured memory / L2 scene block | L3 user persona `persona.md` | Next time you show up, the Agent knows you better |
This lets the Agent work the way humans do: read the table of contents first, then chapters, and only crack open the raw material when needed. The context window stops being a desk that just keeps piling up — it becomes a workspace you can fold and unfold.
Long-term memory is more than "save the chat log somewhere". The real value is mining stable preferences, implicit goals, and contextualized experience out of conversational fragments.
The shape mirrors a DIKW pyramid: from Data to Information to Knowledge to Wisdom. The Agent stops merely recalling "what the user said" and starts understanding "what the user might need".
The biggest risk of compression is "saving tokens but losing the receipts". So TencentDB Agent Memory does not collapse history into an irreversible summary — it keeps a clear path from high-level abstraction back to ground-truth evidence.
The upper layer carries "judgment" and direction; the lower layer carries "evidence" and precision. Short-term compression and long-term memory close into one loop: **foldable and unfoldable; abstract yet auditable.**
Many memory systems break here: when recall is wrong, all you see is a list of vector scores, and you cannot tell where things went sideways. TencentDB Agent Memory keeps the key intermediates as readable files:
Debugging is no longer rummaging through a black-box database — it is walking the chain "persona → scene → memory → raw text" until the issue surfaces.
Long-term memory and short-term compression look like two features but follow the same storage principle underneath: **the database stores searchable facts; the file system stores readable, editable structure.**
The lower stack is the armory: stable, complete, retrievable. The upper stack is the battle map: flexible, readable, fast to iterate. The short-term canvas and the long-term persona are both, at heart, **high-density working surfaces the Agent can read.**
We welcome every kind of contribution — bug reports, feature ideas, doc fixes, benchmark reproductions, ecosystem integrations, or a Pull Request. Agent memory is far from a solved problem, and we'd love to figure it out together.
- 🐞 **Found a bug or have a question?** Open an issue at [GitHub Issues](https://github.com/Tencent/TencentDB-Agent-Memory/issues) — we respond within 24 hours.
- 💡 **Have an idea to share?** Start a thread in [GitHub Discussions](https://github.com/Tencent/TencentDB-Agent-Memory/discussions).
