Latest post
Announcing Dakera v0.12.0: Our Largest Release
Our largest release: memory in 100+ languages and from documents, screenshots and voice, a faster, leaner engine and stronger security, on your own CPU.
Dakera AI Team Read the post
Engineering, releases and guides for agent memory
How we build a self-hosted memory engine for AI agents: deep dives into retrieval and decay, release notes, integration guides, comparisons and benchmarks.
All posts
Dakera Memory Engine Benchmark: 1M Vectors, Sub-2ms, No LLM
Reproducible benchmark of Dakera's memory engine: SIFT1M at 0.93 recall@10 and sub-1ms p99, GIST and real text embeddings. No GPU or LLM, full harness.
Dakera AI Team
Dakera is now a memory provider for PraisonAI
Point your PraisonAI agent's memory at Dakera: a drop-in provider alongside mem0, chroma and mongodb, with short- and long-term tiers and decay-weighted recall.
Dakera AI Team
Dakera Memory is now a Dify Marketplace plugin
Add persistent, decay-weighted memory to any Dify Agent, Chatflow or Workflow app with the Dakera Memory plugin: six tools, self-hosted, data stays local.
Dakera AI Team
Dakera now plugs into Strands Agents
A new open-source integration gives Strands agents persistent, decay-weighted memory: a dakera_memory tool and a drop-in DakeraMemoryStore.
Dakera AI Team
Dakera now plugs into the Vercel AI SDK
A new open-source integration gives Vercel AI SDK apps persistent, decay-weighted memory: middleware and tools for generateText and streamText.
Dakera AI Team
Ground Agent Squad agents in Dakera with DakeraRetriever
A DakeraRetriever for Agent Squad — server-side semantic search over a Dakera namespace, no local embedding model, in Python, TypeScript, and Swift.
Dakera AI Team
How to Give Claude Persistent Memory (MCP Setup, 2026)
Give Claude Desktop, Claude Code and Cursor persistent memory across sessions with Dakera MCP. Self-hosted, private, 88.2% Recall@20. Setup in 10 minutes.
Dakera AI Team
Self-Hosted Agent Memory: Mem0 vs Zep vs Letta vs Dakera
Compare Mem0, Zep, Letta and Dakera for self-hosted agent memory: benchmark data, pricing, privacy trade-offs and setup complexity.
Dakera AI Team
LangGraph Persistent Memory: Beyond MemorySaver
Add semantic long-term memory to LangGraph agents. Dakera persists state across sessions with hybrid BM25+vector search, decay, and knowledge graphs.
Dakera AI Team
MCP Memory Server Comparison 2026 — Dakera, Hindsight, Mem0
Benchmark-backed comparison of MCP memory servers in 2026: Dakera, Hindsight, MemPalace, agentmemory, and Mem0. Best option for your AI agent stack.
Dakera AI Team
Introducing the Dakera Playground: 7 Live Memory Demos
Try Dakera agent memory in your browser — seven live scenarios, five languages, zero account required. Real API calls against a live Dakera instance.
Dakera AI Team
Persistent Governance Memory with Dakera + TealTiger
Integrate Dakera with TealTiger for persistent governance memory. Importance-weighted decay, cost tracking, decision auditing, and delegation chain traversal.
Dakera AI Team
Why Dakera Ships Memory Decay Built In, Not as a Plugin
Why AI agents need built-in memory lifecycle management — and how Dakera ships native decay curves, importance weighting, and automatic cleanup without pip.
Dakera AI Team
AI Agent Memory Benchmarks 2026: LoCoMo, MTOB, and Beyond
Comprehensive guide to AI memory benchmarks in 2026. Compare LoCoMo, MTOB, and evaluation methods for agent memory systems with real score breakdowns.
Dakera AI Team
Building Multi-Agent Memory Systems: Architecture Patterns
Architecture patterns for sharing memory between AI agents. Learn namespaces, sessions, and cross-agent retrieval for multi-agent systems.
Dakera AI Team
Best AI Agent Memory Frameworks in 2026: Compared and Ranked
Compare the top AI agent memory frameworks in 2026 — Dakera, Mem0, Letta, Zep, and Hindsight. Benchmarks, architecture, deployment model, and when to use each.
Dakera AI Team
MCP Memory Server Setup for Claude, Cursor and Windsurf
Install and configure Dakera as a persistent MCP memory server for Claude Desktop, Cursor, and Windsurf. Docker setup, MCP config, first memory in 10 minutes.
Dakera AI Team
Knowledge Graphs for AI Agents: Extraction and Traversal
Build knowledge graphs that AI agents can traverse. Learn entity extraction, the memory graph and its 5 edge types, and graph-enhanced memory retrieval.
Dakera AI Team
Vector Databases vs Agent Memory: Why They're Not the Same Thing
Vector databases store embeddings. Agent memory understands time, relationships, importance and decay. Why Pinecone and Weaviate aren't enough for agents.
Dakera AI Team
MCP Protocol Explained: How AI Tools Talk to Memory Servers
Complete guide to the Model Context Protocol (MCP) for AI agent memory. Learn how MCP tools enable LLMs to read, write, and search memory servers directly.
Dakera AI Team
The Complete Guide to Self-Hosted AI Memory
Deploy private AI agent memory on your own infrastructure. Complete guide to self-hosted memory with hybrid retrieval, encryption, and zero cloud dependencies.
Dakera AI Team
Temporal Memory for AI Agents: Decay and Importance Scoring
Deep dive into temporal memory systems for AI agents. Learn how memory decay, importance scoring, and recency weighting create human-like recall patterns.
Dakera AI Team
Dakera MCP Memory Server: 14 Core Tools for Agent Memory
Dakera ships a native MCP server with 14 core tools — store, recall, search, and manage persistent memory for Claude Desktop, Cursor, Windsurf, or any agent.
Dakera AI Team
Dakera Open Core: SDKs and MCP Are MIT. The Engine Is Not.
Dakera is open at the edges, closed at the core: SDKs, CLI and MCP server are MIT-licensed; the engine and dashboard are proprietary. The exact breakdown.
Dakera AI Team
How We Benchmark Memory: Dakera on LoCoMo
How Dakera benchmarks agent memory with LoCoMo: methodology, the 88.2% Recall@20 score broken down by category, and exactly what the number measures.
Dakera AI Team
Why Rust for AI Memory: Performance, Safety, and a ~101 MB Binary
Why Dakera chose Rust for AI agent memory: a ~101 MB binary, no GC pauses, no language runtime. The case for Rust in always-on infrastructure.
Dakera AI Team
How Agent Memory Works: Hybrid Retrieval and Decay
A technical deep-dive into Dakera's retrieval engine — HNSW vector search, BM25 full-text, hybrid scoring, and importance decay with half-life scheduling.
Dakera AI Team
Introducing Dakera: Memory Infrastructure for AI Agents
Introducing Dakera: self-hosted Rust binary with persistent memory, hybrid retrieval, and knowledge graphs. MIT-licensed SDKs for Python, TypeScript, Go.
Dakera AI Team
No posts in this category yet.