Blog
Engineering and product posts from the cachly team — persistent AI memory, causal_trace, brain_from_git, MCP server setup, and developer infrastructure.
Your pipeline deleted its own alarm (two greps to check)
A reader said my guard only proved a promise in source control. One run on the real path proved him right: a daily message had been missing for two days.
Your linter has been spreading the bug it was built to catch
One number appeared with five different values across a codebase. The tool meant to prevent that was turning correct values into wrong ones. Here is the fix.
Your cron job has never run. Its 33 tests still pass.
A monitoring tool passed 33 checks and never ran once, because nothing scheduled it. The question missing from many acceptance criteria, and a guard for it.
Your CI is not flaky. Your cache expires every seven days.
A lint job reported zero issues and failed anyway: 311 seconds against a 300 second limit, on a cold cache only. How to tell a real flake from a fixed failure.
Green publish pipeline, and nobody can install your plugin
A publish job reported success for three weeks while the plugin was unreachable in the marketplace. Here is the check that catches it.
Multi-Tenant AI Memory: One Brain and One Key per Customer
How a SaaS app gives each end customer an isolated AI memory on cachly: one instance and one API key per customer, so a leaked key reaches one customer.
Your real ROI in the IDE: cost per call and weekly trend
cachly's IDE plugins now read your configured cost per call and show a week-over-week trend, so the ROI in your status panel matches your actual bill.
causal_trace: Your AI Already Knows Why This Broke
cachly's causal_trace checks your Brain first: if this error, or one like it, was solved before, you get the known fix right away. Here is how it works.
OpenClaw + cachly: One Brain Across Every AI Channel
The @cachly-dev/openclaw Brain Bridge lets your OpenClaw agents share one lesson store with Claude Code, Cursor and the IDE plugins, across all channels.
smart_recall: Brief Your AI Before It Starts
cachly's smart_recall checks what your Brain already knows before your AI starts a task, so relevant lessons, past failures and shortcuts arrive first.
Team Brain: Shared AI Memory for Your Engineering Team
Team Brain makes AI memory compound for the whole team: every fix, gotcha and architecture decision is captured once and ready in whatever AI tool they use.
The Compounding Brain: Why AI Memory Grows in Value
AI memory compounds. The story behind cachly, and why a Brain that keeps every fix, revert and decision grows in value for your team with every week.
PIIGuard: Mask PII Before It Reaches the Server
PIIGuard masks emails, phone numbers, IBANs, IPs and custom secrets on your machine. The cachly server only sees deterministic [MASKED_xxxxxx] tokens.
Privacy-First AI Memory for Fintech & Healthcare Teams
Persistent AI memory for regulated teams: cachly combines client-side PIIGuard masking, EU-only servers and Team Brain, so sensitive data stays on your side.
Same Brain, Any Model: Claude, Cursor, Copilot, Windsurf
Keep your AI memory when you switch models: cachly gives Claude Code, Cursor, Windsurf, Copilot, Cline, Zed and Continue the same Brain.
brain_predict: Predict Failures Before They Reach Production
brain_predict checks your causal knowledge graph before every deploy and returns weighted warnings about what is likely to break, based on your own incidents.
Cachly vs MemGPT: Which AI Memory Is Right for Developers?
MemGPT is now Letta: an agent with its own memory. cachly gives memory to the coding assistant you already use. How they differ, with sources.
Cursor AI Persistent Memory: Keep Context Across Sessions
cachly gives Cursor persistent memory over MCP: it keeps your stack, your fixes and your architecture decisions across sessions. Set up with one command.
cachly is now on awesome-mcp-servers
cachly is now listed in punkpeye/awesome-mcp-servers on GitHub. The story behind it: Team Brain for organisations, CI-driven learning and gRPC for AI agents.
Many npm installs, one real user: both true at once
Many npm installs and exactly one real user. Why both are true, what that says about the path, and the one promise we make to every human.
What worked for our MCP memory server on npm in 2026
What we did to grow cachly's MCP server on npm, what we measured, and what we learned from our first real signups.
Cachly vs CLAUDE.md: A CLAUDE.md Alternative That Scales
CLAUDE.md is a good starting point for AI memory. cachly is the next step: persistent memory over MCP, learning from git, and failure prediction.
Claude Code Persistent Memory: Explain Your Stack Once
Claude Code starts each session fresh. Give it persistent memory with cachly's MCP server: it detects your setup and learns from your git history.
How to Give Cline Persistent Memory with MCP
Add cachly's MCP server to give Cline persistent memory across sessions. It learns from your git history and shares what it knows across your editors.
Windsurf Persistent Memory: Explain Your Stack Once
Windsurf starts each session fresh. Give it persistent memory with cachly's MCP server: it detects your setup, learns from git history, shares with your team.
Zed Persistent Memory: Explain Your Stack Once
Zed starts each session fresh. Give it persistent memory with cachly's MCP server: it detects your setup, learns from git history and shares with your team.
Cachly vs mem0: A mem0 Alternative for Developers
mem0 builds memory into AI agents and apps; cachly keeps what coding agents learn. Both work in Claude Code. How they differ, with a source for each fact.
How to Give GitHub Copilot Persistent Memory with MCP
GitHub Copilot starts each session fresh. Give it persistent memory with cachly's MCP server. Works in VS Code and JetBrains.
AI Memory for Arabic Developers: RTL, MENA Region, GDPR
How developers in Dubai, Riyadh, Cairo and across MENA use cachly: native Arabic RTL support, EU servers under GDPR, and memory for Claude, Cursor and Copilot.
AI Memory for Indian Developers: Hindi, Bengali, Tamil
How developers in Bengaluru, Mumbai and Hyderabad use cachly: support for Hindi, Bengali, Tamil and more Indian languages, DPDP compliance, persistent memory.
AI Memory for Israeli Developers: Hebrew RTL and GDPR
How developers in Tel Aviv and across Silicon Wadi use cachly: native Hebrew RTL support, persistent memory for Claude and Cursor, EU data handling under GDPR.
AI Memory for Russian-Speaking Developers: Cyrillic, GDPR
How Russian, Ukrainian and Bulgarian developers use cachly: native Cyrillic support, persistent memory for Claude and Cursor, and servers in the EU under GDPR.
MCP as a distribution channel, and our pivot to AI memory
How @cachly-dev/mcp-server grew through organic npm installs, why we moved from semantic cache to AI memory, and why MCP works as a channel for dev tools.
Causal Knowledge Graph: how cachly learns cause and effect
cachly stores cause and effect: what led to what, with Bayesian confidence scores. The Causal Knowledge Graph powers brain_predict, ckg_inspect and recall.
Your AI arrives pre-briefed: fully automatic Brain sessions
Since cachly MCP 0.10.9, sessions start automatically when your editor connects. The Brain starts on ListTools and adds context to the first response.
VS Code offline lesson queue: lessons saved while offline
cachly Brain 0.7.8 for VS Code queues lessons locally while the Brain is unreachable and syncs them on reconnect. Works on flights and during maintenance.
Watch Your AI Brain Grow in 3D — Live
A live 3D view of your AI assistant's Brain: every lesson, fix and pattern becomes a glowing node in a real-time galaxy that grows as you work.
Introducing the Cognitive Cache: The Cache That Thinks
cachly v0.6 adds five capabilities: causal root-cause tracing, knowledge decay scoring, brain diff, memory consolidation and zero-config autopilot.
God View: See Every AI Brain in Your Organization at Once
A Global Brain admin view that brings every AI assistant's Brain in your organization into one 3D galaxy, with anomaly detection and cross-team learning paths.
brain_from_git: Bootstrap AI Memory from Your Git History
Your git log holds years of team knowledge: bugs you fixed, approaches that failed, files that keep breaking. brain_from_git turns it into AI Brain lessons.
Zero AI Onboarding: Full Codebase Context on Day One
With brain_from_git and cachly, a new developer's AI assistant gets your team's knowledge in one command: the new developer runs join, and is up to speed.
Your AI Brain Now Speaks Arabic and Hebrew
cachly v0.5.48 adds Arabic and Hebrew: RTL tokenization, Arabic light stemming, stopword lists and cross-language retrieval with the full synonym graph.
Persistent Memory for AI Agents: LangChain, AutoGen, CrewAI
Give AI agents memory that survives restarts, scale-out and multi-run workflows. Adapters for LangChain, AutoGen, CrewAI and OpenClaw, with semantic search.
Brain Levels & Weekly Digest: Your AI's Progress Report
Brain Levels show your AI's learning: every 10 lessons earn a level. The Weekly Digest sends a summary to Slack, email or Telegram: what your AI learned.
Chinese, Japanese & Korean AI Memory: How cachly Handles CJK
Chinese, Japanese and Korean need extra work in many AI memory systems: tokenizers, embedding models, indexing. cachly's key-based memory handles CJK as is.
Store in Japanese, Recall in English: Cross-Language Memory
Store lessons in Japanese, Korean, Arabic or Hebrew and recall them in English, without embeddings. cachly's cross-language retrieval uses a synonym map.
Ambient Git Learning: Your Commit History as AI Knowledge
Your git log holds much of what your team knows. Ambient Git Learning turns commit history into searchable AI Brain lessons, automatically, as you commit.
Memory Crystals: Distilling Team Knowledge into AI Context
Memory Crystals distill what your team has learned into a dense, current snapshot that goes into every AI session automatically.
Team Telepathy: How We Built a Shared AI Brain for Teams
When one developer fixes a hard bug, the whole team benefits. Team Telepathy is a shared AI Brain that brings each lesson to every AI assistant on the team.
Free, Private Embeddings for Your AI Dev Brain with Ollama
We run nomic-embed-text via Ollama on our own servers in Germany: semantic search for every cachly Brain, with no OpenAI API key and no extra cost.
Cachly SDK Integrations: Python, TypeScript, Go, LangChain
How to wire cachly into LangChain, the Vercel AI SDK, OpenAI clients and other stacks. Copy-paste examples for each major language, in a few lines of code.
Self-Host a Semantic LLM Cache in 5 Minutes
Run cachly on your own server with one docker compose command. Your data stays in your infrastructure, air-gapped if you like, and GDPR-ready.
AI memory for your coding assistant, without an OpenAI key
We made AI memory work without any API key or embedding provider. Before: your assistant forgot everything between sessions. After: it remembers in instant.
Your AI Brain in VS Code and IntelliJ: cachly IDE Plugins
New IDE plugins show brain health, lesson count, and recall stats directly in your status bar. VS Code and IntelliJ/WebStorm — zero config.
Building an IntelliJ Plugin in Kotlin: A Status Bar Widget
From build.gradle.kts to a live status bar widget in IntelliJ IDEA, WebStorm and other JetBrains IDEs: the full Kotlin walkthrough, with the tricky parts.
How we cut LLM costs with Semantic Cache
Users rephrase the same question in many ways. Semantic caching with pgvector similarity search answers repeats from cache. With 3 lines of code.
Building a VS Code Extension That Shows What My AI Learned
A step-by-step guide to building a VS Code status bar extension — from `yo code` to a live widget showing your AI brain's lessons and token savings.
We built persistent memory for Claude Code
How we gave coding assistants a brain that lasts across sessions: session briefings, lesson recall, team knowledge and semantic search on Valkey and pgvector.
Search Your Japanese AI Brain in Romaji
cachly v0.5.37 adds Hepburn romaji search for katakana: type 'kontena' and find コンテナ lessons instantly. Digraphs, geminates, and long vowels all handled.
Your AI Remembers: Introducing the cachly Brain IDE Plugins
VS Code and IntelliJ plugins that show what your AI coding assistant has learned: lesson count, recall stats and estimated token savings in the status bar.