agent-memory
One Bank, Two Audiences, Two Briefs
Hindsight 0.10.2 lets each observation scope consolidate under its own mission. The company-wide view can record trends while a user's own scope keeps the detail, from the same facts.
I Gave Meta's AI Assistant My Entire Memory. Here's What It Found.
We connected Meta Muse to a live Hindsight account and asked what it knew. It runs in Meta's cloud with no plugin host, so there is nothing to install: a prompt, an OAuth sign-in, and it reads the banks your other AI tools already write to.
40,000 Stars, and the Number We Actually Watch
Hindsight just crossed 40,000 GitHub stars in eleven months. The stats behind it, the 45 days it took to double, and the one number that tells us more than stars do.
We Added Jev as a Reranker. Here's What We Learned.
Six things we learned adding Jev, a model that returns typed decisions instead of text, as a reranker in Hindsight 0.10.1. Including two designs that failed and the numbers that killed them.
Bring the Facts, Not the Beliefs
When you migrate a memory history, its duplicate conclusions come with it. The fix is to import the source material and let your bank rebuild the beliefs itself.
A Bank Is Now Something You Can Pick Up
Clone a memory bank in one call, export and import it with an explicit scope, or rename it in place. What travels, what gets rebuilt, and where it will bite you.
What Hindsight Learned This Summer
Six releases in eight weeks, from 0.8.5 to 0.10.0. The biggest things Hindsight learned to do this summer, grouped by what they let you build.
See the Exact Prompt Before You Spend a Token
You configured a mission, a disposition and three strategies. What did the model actually receive? Hindsight 0.10.0 will show you, without spending a token.
What It Takes to Put a Screenshot in an Agent's Memory
Storing an image is easy. Making a fact cite the screenshot it was read from, without inventing evidence, is the hard part. How attachments work in Hindsight 0.10.0.
Stop Growing Your Always-On Context
The default fix for an agent that forgets is to put more in the prompt. It works until it doesn't, and the way it fails is quiet: unconditional context that costs every turn and gets less relevant as it grows.
Onboarding an Engineer vs. Onboarding an Agent
You already know how to get a new engineer productive: docs, pairing, and time. Agents need the same thing and get none of it, because every session is their first day.
Enterprise Features in Hindsight Cloud
Every enterprise capability in Hindsight Cloud, and the failure each one prevents: single sign-on, MFA enforcement, role-based access, scoped API keys, audit logging, Memory Defense, webhooks, and bring-your-own-cloud or on-premise deployment.
Cross-Encoder Reranking: The Last Stage of Agent Memory Recall
Retrieval hands the reranker a few hundred plausible memories. What happens next decides what your agent actually sees. Inside Hindsight's cross-encoder stage: score normalization, multiplicative boosts, candidate budgets, and how to make recall fail open.
Your Agent Can Now Manage Its Own Knowledge Base Over MCP
Hindsight 0.9.2 turns the knowledge base into a native MCP surface: seven agent-facing tools to browse, search, create, update, and prune its own wiki, over the same connection it already uses to retain and recall.
Knowledge Graphs vs. Vector Search for Agent Memory
Should your agent's memory be a vector store or a knowledge graph? They solve different halves of the problem. Here is where each one wins, where each breaks, and why agent memory needs both.
Give Your elizaOS Agent a Long-Term Memory
elizaOS agents run around the clock and talk to everyone. The Hindsight plugin gives them persistent, per-user long-term memory: recall before each model call, retain after each turn, fail-safe by default.
Introducing Hindsight Academy: Learn Agent Memory by Doing
Hindsight Academy is a free, hands-on course that teaches agent memory by doing: from why agents forget to a production deployment, in four courses and about an hour.
Give Every Hermes Bot Its Own Memory
Nous just shipped Bot Mode for Hermes: named bots, each with its own memory. With Hindsight, that memory is persistent and precisely scoped per bot.
Give DeepSeek Harness a Memory of Your Codebase
DeepSeek just launched Harness (dsh), an open-source coding agent built on plugins. Hindsight plugs in natively to give it long-term memory of your codebase.
20,000 Stars: How Hindsight Got Here, Version by Version
Hindsight just crossed 20,000 GitHub stars in under ten months and 67 releases. Here's the timeline: every major capability we shipped, version by version.
Claude Code Now Builds and Reads Its Own Knowledge Base
With Hindsight's Coding Agents integration, Claude Code surveys your repo, builds a self-healing wiki of it, keeps it current, and reads it before every task.
Give Any Agent Plugins Client Long-Term Memory
Agent Plugins is the new vendor-neutral standard for packaging Skills and MCP servers. Hindsight ships one portable plugin that gives Codex, Cursor, Copilot, Kiro, and VS Code long-term memory.
Best Open-Source Agent Memory Systems (Self-Hosted, 2026)
A benchmark-grounded, fair comparison of the open-source agent memory systems you can self-host: Hindsight, Mem0, Graphiti/Zep, Letta, and Cognee.
Reasoning Isn't the Hard Part Anymore. Continuity Is.
How writenode uses Hindsight — a per-user memory bank plus a mode classifier — to carry a person's thinking forward across sessions instead of starting from a blank page.
Per-User Memory for AI Products: Multi-Tenant Patterns
How to give every user isolated memory in a multi-tenant AI product: hard boundaries vs soft partitions, testing for cross-tenant leakage, GDPR deletion, and scaling.
The 10 Things to Look For in an Agent-Memory System
The 10 things that separate a real agent-memory system from a demo: retrieval, conflict resolution, freshness, security, self-hosting, observability, and more.
Give GitHub Copilot CLI a Memory of Your Codebase
GitHub Copilot CLI forgets your project between sessions. Add Hindsight so it recalls past decisions and retains what it learns, automatically via hooks.
How to Move Your Agent's Memory Off a Vector Database
A practical guide to migrating agent memory off a vector database like Pinecone or Chroma to Hindsight: what to export, how to map it, and what you gain.
8 Agent Memory Use Cases (With Real Examples)
The real agent memory use cases teams build on Hindsight, from coding agents to customer-facing products to automations, each with a concrete example.
recall vs reflect: Search Your Agent's Memory, or Ask It
Hindsight gives your agent two ways to read its memory. recall retrieves the relevant facts (no LLM, instant). reflect reasons across them to synthesize an answer. When to use each.
Give Roo Code a Memory So Every Task Builds on the Last
Roo Code runs autonomous coding tasks but starts each one from zero. Add Hindsight agent memory: recall context before every task, retain a summary after.
Your 1M-Token Context Window Is Not Memory
A context window is working memory: it forgets everything when the session ends and degrades long before it fills. Why a bigger window is still not agent memory.
Mira Murati's Thinking Machines Dropped Inkling. We Made It Your Agent's Memory in 5 Minutes.
Thinking Machines open-sourced Inkling. We made it an AI agent's long-term memory in four env vars, and it resolved 'last week' to a real date on the first try. Here is how.
Give ZCode a Long-Term Memory That Follows You Between Tools
ZCode is Z.ai's GLM desktop coding agent. Add Hindsight for persistent memory via plain Python hooks, no MCP, shared with Claude Code and Cursor.
The Fully Open Agent Memory Stack: Self-Hosting Hermes + Hindsight
Most agent-plus-memory stacks have a closed core: a frontier API model, a proprietary memory service, or both. Hermes and Hindsight are the exception. Every layer is open source and can run on your own hardware with no external calls.
One Bank or Many? A Field Guide to Structuring Agent Memory
Every Hindsight integration tells you to set a bank_id, but not how to decide what a bank should be. A bank is a recall boundary: how to scope agent memory, and when to reach for tags instead.
Give Every Agent You Run in Omnigent a Persistent Memory
Omnigent wraps any AI agent: Claude Code, Codex, Cursor, Hermes, and more. Add Hindsight once and every harness it orchestrates gets recall, retain, and reflect from a single setup.
Inside retain(): What Actually Happens When Your Agent Remembers
Calling retain() is one line of code. Underneath, a sentence becomes extracted facts, resolved entities, a knowledge graph, and consolidated observations. Here is the whole write path.
One Memory, Three Surfaces: A Day With Cursor, OpenClaw, and Vapi
We pointed our code editor, our Slack agent, and our voice agent at one shared Hindsight memory. Here is a day in that workflow, and what actually changed.
Entity Resolution in Agent Memory: One Person, Many Names
How agent memory decides that Sarah, Sarah Chen, and she are one person, and why Hindsight resolves entities with a co-occurrence graph, not embeddings.