teolex2020/AuraSDK

Aura turns fragile prompt-only agents into auditable, memory-aware, production-ready systems
Deterministic · No fine-tuning · No cloud training · Local recall · No required embeddings
  
  
---
Your AI model is smart. But it forgets everything after every conversation.
> **New: Aura for Windows** — a free desktop app built on this library that gives Claude Code, Cursor, Codex, Copilot and other AI tools one shared local memory.
Aura is a local cognitive runtime that runs alongside any frozen model. It gives agents durable memory, explainability, governed correction, bounded recall reranking, and bounded self-adaptation through experience — all locally, without fine-tuning or cloud training.
No API keys. No embeddings required. No cloud. The model stays the same — the cognitive layer becomes more structured, more inspectable, and more useful over time.
What's new in 1.61
- **Your words stay apart from outside text.** recall() now returns the provenance context by default: first-hand memory first, then web, document and tool content fenced and quoted as data. The fence stops ordinary instructions hidden in that content; text written to pose as the user can still steer a model (experiments/mcp_gateway_fence_robustness, experiments/agentdojo_fence). **Upgrade note:** code that parses the old === COGNITIVE CONTEXT === text should pass format="levels".
- **Security profiles.** Aura(path, security="strict"), set_security_profile() and security_report(). Memories written by a model through MCP are stored as inferred, not as the user's word. One untrusted source repeating a claim can no longer crowd out a first-hand fact.
- **Better recall context.** A short block of lasting facts about the user, event dates on first-hand memory, and relevance-trimmed context.
- **One memory for many AI tools.** The Rust features mcp-http and mcp-bridge add a token-protected loopback MCP server and the aura-bridge stdio bridge, so several clients share one store. This is what the Aura for Windows app runs on.
- **Conversation capture.** python -m aura capture is a Claude Code hook that keeps the user's words as first-hand memory.
- **Safer HTTP server.** python -m aura serve refuses to listen beyond localhost without an API key. Install it with pip install "aura-memory[http]".
- **Plain MIT.** The license-enforcement module is gone.
The full list is in the changelog.
> **⭐ If Aura is useful to you, a GitHub star helps us get funding to continue development from Kyiv.**
---
Why Aura?
| Property | Aura |
|---|---|
| Architecture | Five-layer cognitive engine, with Belief→Concept→Causal→Policy derivation |
| Memory operations | Local Rust computation; no required LLM or embedding API |
| Recall latency | 11.61 ms median, distinct uncached top-10 queries on 5K records; see Performance |
| Offline use | Supported without a cloud service |
| API fee for memory operations | None required; local compute and storage still have costs |
| Package size | 2.77 MB Windows CPython 3.13 wheel 1 |
| Lifecycle and governance | Built-in decay/promotion, provenance, contradiction controls, purge/freeze |
| Encryption at rest | Optional ChaCha20-Poly1305 with Argon2id key derivation |
We do not publish unmeasured speed, cost, or feature rankings against other
memory products. Their deployment modes and search outputs differ; a meaningful
comparison needs matched data, workload, and quality criteria.
The Core Idea: Cheap Model + Aura > Expensive Model Alone
Fine-tuning costs thousands of dollars and weeks of work. RAG requires embeddings and a vector database. Context windows are expensive per token.
Aura gives you a third path: **a local cognitive runtime that accumulates structured experience between conversations** — free and local.
The model stays the same. The cognitive layer gets stronger. That's Aura.
Performance
**Current local measurement (Aura 1.60.0).** On Windows 10, AMD Ryzen 5
5600X, and CPython 3.13.14, we stored the first 5,000 dialogue turns from
LoCoMo (median 154 bytes per turn). Recall used 100 distinct natural-language
questions, recall_structured(top_k=10), and no repeated-query cache hits.
One sequential local run, measured on 2026-09-25:
| Operation | Median | P95 |
|-----------|-------:|----:|
| Store one turn | 5.66 ms | 10.10 ms |
| Structured recall, distinct uncached query | 11.61 ms | 18.95 ms |
| Get record by known ID | 0.0033 ms | 0.0083 ms |
Local MCP stdio added about 0.04 ms to mean recall time in a separate