Build your second brainwith one line of code.
Add memory and learning to your existing LLM client. Manage memories, recipes and feedback through dedicated APIs.
# Your existing client
client = openai()
# Add one line
client = cosens(client)How it works
Your agent reasons with an LLM, recalls relevant context from Memory, and uses Learn to improve its methods. Follow a task through that process.
You ask.
Start with the question you actually need help with.
Swipe to explore the diagram →
SDK / Learn preview · Scripted scenario. Method validation is shown as a process, not a completed live evaluation.
Add one line.
Take control
with APIs.
Create your LLM client as usual, then wrap it with CoSens. The proposed SDK also includes Memory, Recipe and feedback APIs for managing what your agent remembers and the methods it draws on.
Memory, compared.
Selected benchmark scores for CoSens and the compared memory systems. Higher is better.
4 methods
| Benchmark | CoSens | Mem0 | A-Mem | MemGPT |
|---|---|---|---|---|
| LoCoMoOverall Token-level F1 · 0–1; higher is betterLatest Memory comparison table | 0.368945 | 0.256981 | — | — |
| LoCoMo-PlusFactual average · 0–100; higher is betterEarlier consolidated report · paper references and company result | 61.52 | 57.24 | 59.64 | — |
| HaluMemMedium QA Correct · 0–1; higher is betterLatest Memory comparison table | 0.537369 | 0.492279 | — | — |
| LongMemEvalEvidence Recall@10 (%) · 0–100; higher is betterEarlier consolidated report · retrieval evidence, not answer accuracy | 95.2 | — | — | — |
Experimental results and source-report references. Each row names its metric, scale and source. — means no result is available. Settings may differ between systems.
Experimental results are provided for reference only and do not constitute academic research. Results may be updated at any time as new evaluations become available.
More benchmark... ↗Your agent.
Your experience.
Add the experience layer to your product.