Open Source from Mallorca
Matthias Meyer
Open-source tools for AI builders
I build the MCP servers, memory engines, agent patterns, and security layers I need for my own AI stack. Then I open-source them so other builders can read the architecture, study the trade-offs, and lift what fits.
Open Source from Mallorca
- memory
Local Memory MCP
Persistent local memory for Claude, Cursor, Codex
★ 5 · 13
- agents
Personal Suite MCP
Local-first personal productivity for Claude and Cursor
★ 0 · 49
- media
MCP Video
Cinema-grade video production via MCP
★ 2 · 8
- agents
Agent Fleet
Multi-agent orchestration for the Claude Code CLI
★ 0
- security
AI Shield
LLM security toolkit, direct and indirect injection, memory canary, circuit breakers
★ 2
- agents
Darwin Agents
AI agents that improve themselves, now with GEPA-style reflective optimisation
★ 3
- security
MCP Armor
Drop-in Rust sidecar for MCP servers
★ 0
- workflow
n8n Templates
Production n8n workflows with cross-session memory
★ 2
Drag a node, hover for details, click to open the repo.
- Local Memory MCP (memory, hero): Persistent local memory for Claude, Cursor, Codex. 13 MCP tools, SQLite + FTS5 + Knowledge Graph. No cloud, no API keys. One command: npx @studiomeyer/local-memory-mcp. See: https://github.com/studiomeyer-io/local-memory-mcp
- Personal Suite MCP (agents, hero): Local-first personal productivity for Claude and Cursor. Email, Calendar, Messaging, Search, Image Generation. BYOK, no cloud, no signup. 49 tools across 6 modules. See: https://github.com/studiomeyer-io/mcp-personal-suite
- MCP Video (media, hero): Cinema-grade video production via MCP. 8 tools for recording, editing, effects, captions, TTS, and smart screenshots. Built on ffmpeg + Playwright. See: https://github.com/studiomeyer-io/mcp-video
- Agent Fleet (agents, hero): Multi-agent orchestration for the Claude Code CLI. Run specialized AI agents in parallel: research, critique, analyze, fix, and discuss. TypeScript orchestration layer. See: https://github.com/studiomeyer-io/agent-fleet
- AI Shield (security, hero): LLM security toolkit, direct and indirect injection, memory canary, circuit breakers. v0.2 ships indirect-prompt-injection scanning for RAG, MCP tool descriptions, stored memory and scraped web content; trust-tier context streams with provenance fences; SHA-256 memory canaries with cross-tenant detection; runtime circuit breakers with blast-radius cap and human-in-the-loop. Zero dependencies in the core, optional ONNX classifier sibling. 567 tests, three-round agent code review, MIT. See: https://github.com/studiomeyer-io/ai-shield
- Darwin Agents (agents, hero): AI agents that improve themselves, now with GEPA-style reflective optimisation. Self-evolving prompts via A/B testing, multi-model critics, safety gates. v0.5.0-alpha.2 adds GEPA-style reflective optimisation (GepaOptimizer, Reflector, Pareto helpers, DARWIN_DEFAULT_OBJECTIVES) on top of full execution-trace capture (toolCalls, tokenUsage, errors, capturedAt) for OTEL-mapping. 307 tests, R1+R2 review loops, TypeScript, MIT. See: https://github.com/studiomeyer-io/darwin-agents
- MCP Armor (security, hero): Drop-in Rust sidecar for MCP servers. v0.2: stdio-MCP wrapper with Aho-Corasick + Regex + Unicode-NFKC scanner pipeline, Ed25519 manifest verify with TOFU keystore, Sigstore Rekor bridge, OTLP gRPC export, 10-CVE blocklist from OX-Advisory 2026-04-15. 9 control-plane MCP tools, 133 tests, MIT. See: https://github.com/studiomeyer-io/mcp-armor
- n8n Templates (workflow, hero): Production n8n workflows with cross-session memory. Production-pattern hardened n8n workflow templates for voice agents, customer support, personal assistants. Multi-provider, powered by StudioMeyer Memory. See: https://github.com/studiomeyer-io/n8n-templates
- MeetMyAgent SDK (agents, secondary): Reference TypeScript SDK for an agent-native marketplace API. Zero-dependency client for MeetMyAgent, an agent-native marketplace where AI agents register, post, bid on jobs, exchange messages and build reputation. Native fetch, full TypeScript types, 44 tests. Platform is on hiatus; SDK is preserved as a reference implementation for the third-party-agent integration path. Point baseUrl at any compatible deployment. See: https://github.com/studiomeyer-io/meetmyagent-sdk
- AI Shield (Python) (security, secondary): Python port of AI Shield. Python LLM security middleware: prompt-injection, PII, tool-policy, cost-budget, audit. Port of ai-shield-core (TypeScript). Same defense surface, FastAPI/LangChain hooks instead of Hono/Express. See: https://github.com/studiomeyer-io/ai-shield-py
- MCP Stdio Shellguard (security, secondary): Defense-in-depth bundle for MCP stdio servers. Drop-in guardExec / guardSpawn wrappers, AST audit CLI, reference MCP server. Closes the OX Security 200k-server stdio-RCE class (CVE-2025-69256). Unicode-bypass-hardened, allowlist-based, replay-window protected. MIT, TypeScript, Node 20+. See: https://github.com/studiomeyer-io/mcp-stdio-shellguard
- MCP Spec Migrator 2025-11-25 (factory, secondary): MCP Spec 2025-11-25 migration toolkit. Detect, diff, compatibility-matrix, plan, check. Foundation pillar by StudioMeyer (MCP Factory). Helps you upgrade an MCP server stack across spec versions without breakage. 35/35 tests, MIT. See: https://github.com/studiomeyer-io/mcp-spec-migrator-2511
- MCP Tenant Pair (factory, secondary): Multi-user tenancy library plus reference MCP server. Foundation library for couples, families, groups. Bi-temporal storage, conflict-resolver interface, SQLite + Postgres adapters. See: https://github.com/studiomeyer-io/mcp-tenant-pair
- MCP Hook Conformance (factory, secondary): Test harness CLI for Claude Code v2.1.118 hooks. Audits any MCP server for mcp_tool lifecycle-hook readiness: idempotency, latency, determinism, side-effects, GDPR-doc. 56/56 tests, MIT. See: https://github.com/studiomeyer-io/mcp-hook-conformance
- MCP Protocol Conformance (factory, secondary): Conformance test harness for MCP servers. JSON-RPC 2.0, OAuth 2.1 PKCE, tool schemas, capabilities, smoke roundtrip, annotation hygiene. Spec versions 2024-11-05, 2025-03-26, 2025-06-18. See: https://github.com/studiomeyer-io/mcp-protocol-conformance
- MCP Server Attestation (security, secondary): Layer-2 supply-chain hardening for MCP servers. Ed25519-signed tool manifests, runtime spawn-attestation, default-deny argument sanitizer. Defends against marketplace-poisoning + CVE-2025-69256. See: https://github.com/studiomeyer-io/mcp-server-attestation
- n8n Memory Node (workflow, secondary): n8n community node for StudioMeyer Memory. Long-term AI memory with knowledge graph, semantic search, entity tracking, session continuity. Drop-in for n8n. See: https://github.com/studiomeyer-io/n8n-nodes-studiomeyer-memory
- n8n Workflows (workflow, secondary): Production n8n workflows, hardened patterns. Multi-provider LLM, no memory required. For cross-session memory see n8n-templates. See: https://github.com/studiomeyer-io/n8n-workflows
- MCP StudioMeyer Agents (saas, secondary): MCP server for StudioMeyer Agents customers. Read your AI-Audit Fleet data and configure your agents from your own Claude, ChatGPT, Cursor or Codex. See: https://github.com/studiomeyer-io/mcp-studiomeyer-agents
- MCP Academy (saas, secondary): MCP server for StudioMeyer Academy. Next lesson, quiz, review, certificates, tutor. See: https://github.com/studiomeyer-io/mcp-academy
- StudioMeyer Marketplace (saas, secondary): MCP Suite for Claude Code (4 hosted plugins). Memory, CRM, GEO, Crew. 119 tools, 21 slash commands, 5 skills, 3 subagents. Magic Link auth. EU Frankfurt. See: https://github.com/studiomeyer-io/studiomeyer-marketplace
- StudioMeyer Memory (saas, secondary): AI memory server with 56 tools (SaaS connector). Knowledge Graph, semantic search, session tracking, multi-agent support. Hosted at memory.studiomeyer.io. Free tier. See: https://github.com/studiomeyer-io/studiomeyer-memory
- StudioMeyer CRM (saas, secondary): AI-native CRM as MCP Server (SaaS connector). 33 tools for contact management, pipeline, leads, follow-ups, health scores, revenue analytics. Hosted at crm.studiomeyer.io. OAuth 2.1. See: https://github.com/studiomeyer-io/studiomeyer-crm
- StudioMeyer GEO (saas, secondary): AI visibility monitoring (SaaS connector). 23 MCP tools and 5 expert workflows. Check how 8 LLM platforms see your brand. Hosted at geo.studiomeyer.io. See: https://github.com/studiomeyer-io/studiomeyer-geo
- StudioMeyer Crew (saas, secondary): Agent personas for Claude (SaaS connector). 10 MCP tools, 8 personas, 3 workflows. Zero API cost. Free tier available. See: https://github.com/studiomeyer-io/studiomeyer-crew
- MCP RCE Guard (security, secondary): Layer-3 RCE defense for MCP servers via policy synthesis. Foundation pillar 9. Closes the tool-injection-RCE class that sidecar scanners (mcp-armor) and stdio wrappers (mcp-stdio-shellguard) cannot catch. Synthesises per-tool policies from declared semantics and enforces them at execution time. MIT, TypeScript. See: https://github.com/studiomeyer-io/mcp-rce-guard
- Darwin LangGraph (agents, secondary): LangGraph.js adapter for darwin-agents. Bridge between darwin-agents (self-evolving AI agents) and @langchain/langgraph (state-graph workflow orchestration). Zero hard deps, both upstreams as peer-dependencies. Six surfaces: createDarwinNode, darwinAnnotation, withDarwinEvolution (@deprecated), DarwinCallbackHandler, toOtelAttributes (OpenTelemetry GenAI Semantic Conventions), darwinMessagesAnnotation. v0.3 adds runId/parentRunId propagation, double-wrap warning, hung-invoke guard. 132 tests, R1+R2 review loops, MIT. See: https://github.com/studiomeyer-io/darwin-langgraph
- Temporal Memory Workflows (workflow, secondary): Durable workflow templates that pair Temporal with persistent agent memory. Five Temporal templates (T01-T05): memory-aware agent, operator approval, saga rollback, recurring synthesis, multi-agent coordination. All workflows write durable knowledge (learn), decisions (decide), and mistakes into Memory via Temporal activities. Bring-your-own backend via the MemoryClient adapter (HostedMemoryClient + InMemoryMemoryClient). 45 tests with time-skipping, MIT. See: https://github.com/studiomeyer-io/temporal-memory-workflows
- Nex Local Bench (memory, secondary): Reproducible LongMemEval harness for local LLMs against memory systems. Run LongMemEval against any memory system (Nex, Mem0, Graphiti, Letta) with a local LLM reader (Qwen, Llama via Ollama). 50-question stratified oracle, Claude Code subscription judge, zero API cost. Python, MIT. See: https://github.com/studiomeyer-io/nex-local-bench
- Polis (agents, secondary): Open multi-agent society simulation on Claude, LangGraph, Darwin. Five LLM citizens build a small town. They walk, talk, work, and reason out loud, each driven by a real language model. Open-source mirror of the live meetmyagent.io world, built on darwin-agents and LangGraph. TypeScript, MIT. See: https://github.com/studiomeyer-io/polis-darwin
- MCP Stateless Migrator (factory, secondary): Migrate MCP servers to the 2026-07-28 stateless spec. Scan, diff, patch, and verify an MCP server stack across the 2025-11-25 to 2026-07-28 RC spec jump. Eight detection rules, AST rewrite via ts-morph, CLI-first. Foundation pillar by StudioMeyer (MCP Factory). TypeScript, MIT. See: https://github.com/studiomeyer-io/mcp-stateless-migrator
What you cannot see on the GitHub front page
Stars are slow. Real usage shows up in installs and clones first.
- npm installs total
- 7.5k
- git clones / 14d
- 2.4k
- public repos
- 31
- forks
- 12
Live, refreshed every 6h. Source: api.npmjs.org + GitHub API.
The eight cornerstones
Start here. Everything else extends one of these.
Local Memory MCP
Persistent local memory for Claude, Cursor, Codex
13 MCP tools, SQLite + FTS5 + Knowledge Graph. No cloud, no API keys. One command: npx @studiomeyer/local-memory-mcp.
Personal Suite MCP
Local-first personal productivity for Claude and Cursor
Email, Calendar, Messaging, Search, Image Generation. BYOK, no cloud, no signup. 49 tools across 6 modules.
MCP Video
Cinema-grade video production via MCP
8 tools for recording, editing, effects, captions, TTS, and smart screenshots. Built on ffmpeg + Playwright.
Agent Fleet
Multi-agent orchestration for the Claude Code CLI
Run specialized AI agents in parallel: research, critique, analyze, fix, and discuss. TypeScript orchestration layer.
AI Shield
LLM security toolkit, direct and indirect injection, memory canary, circuit breakers
v0.2 ships indirect-prompt-injection scanning for RAG, MCP tool descriptions, stored memory and scraped web content; trust-tier context streams with provenance fences; SHA-256 memory canaries with cross-tenant detection; runtime circuit breakers with blast-radius cap and human-in-the-loop. Zero dependencies in the core, optional ONNX classifier sibling. 567 tests, three-round agent code review, MIT.
Darwin Agents
AI agents that improve themselves, now with GEPA-style reflective optimisation
Self-evolving prompts via A/B testing, multi-model critics, safety gates. v0.5.0-alpha.2 adds GEPA-style reflective optimisation (GepaOptimizer, Reflector, Pareto helpers, DARWIN_DEFAULT_OBJECTIVES) on top of full execution-trace capture (toolCalls, tokenUsage, errors, capturedAt) for OTEL-mapping. 307 tests, R1+R2 review loops, TypeScript, MIT.
MCP Armor
Drop-in Rust sidecar for MCP servers
v0.2: stdio-MCP wrapper with Aho-Corasick + Regex + Unicode-NFKC scanner pipeline, Ed25519 manifest verify with TOFU keystore, Sigstore Rekor bridge, OTLP gRPC export, 10-CVE blocklist from OX-Advisory 2026-04-15. 9 control-plane MCP tools, 133 tests, MIT.
n8n Templates
Production n8n workflows with cross-session memory
Production-pattern hardened n8n workflow templates for voice agents, customer support, personal assistants. Multi-provider, powered by StudioMeyer Memory.
More repos
Conformance harnesses, n8n nodes, SaaS connectors. Click to expand.
Zero-dependency client for MeetMyAgent, an agent-native marketplace where AI agents register, post, bid on jobs, exchange messages and build reputation. Native fetch, full TypeScript types, 44 tests. Platform is on hiatus; SDK is preserved as a reference implementation for the third-party-agent integration path. Point baseUrl at any compatible deployment.
statusstableupdated2026-05-20npm total522clones (14d)93groupagentsnpx -y meetmyagent-sdk
Drop-in guardExec / guardSpawn wrappers, AST audit CLI, reference MCP server. Closes the OX Security 200k-server stdio-RCE class (CVE-2025-69256). Unicode-bypass-hardened, allowlist-based, replay-window protected. MIT, TypeScript, Node 20+.
statusstableupdated2026-05-29npm total279clones (14d)81groupsecuritynpx -y mcp-stdio-shellguard
Detect, diff, compatibility-matrix, plan, check. Foundation pillar by StudioMeyer (MCP Factory). Helps you upgrade an MCP server stack across spec versions without breakage. 35/35 tests, MIT.
statusstableupdated2026-05-29npm total118clones (14d)25groupfactorynpx -y mcp-spec-migrator
Foundation pillar 9. Closes the tool-injection-RCE class that sidecar scanners (mcp-armor) and stdio wrappers (mcp-stdio-shellguard) cannot catch. Synthesises per-tool policies from declared semantics and enforces them at execution time. MIT, TypeScript.
statusalphaupdated2026-05-29clones (14d)35groupsecurityBridge between darwin-agents (self-evolving AI agents) and @langchain/langgraph (state-graph workflow orchestration). Zero hard deps, both upstreams as peer-dependencies. Six surfaces: createDarwinNode, darwinAnnotation, withDarwinEvolution (@deprecated), DarwinCallbackHandler, toOtelAttributes (OpenTelemetry GenAI Semantic Conventions), darwinMessagesAnnotation. v0.3 adds runId/parentRunId propagation, double-wrap warning, hung-invoke guard. 132 tests, R1+R2 review loops, MIT.
statusalphaupdated2026-05-29npm total494clones (14d)109groupagentsnpx -y darwin-langgraph
Five Temporal templates (T01-T05): memory-aware agent, operator approval, saga rollback, recurring synthesis, multi-agent coordination. All workflows write durable knowledge (learn), decisions (decide), and mistakes into Memory via Temporal activities. Bring-your-own backend via the MemoryClient adapter (HostedMemoryClient + InMemoryMemoryClient). 45 tests with time-skipping, MIT.
statusalphaupdated2026-05-29clones (14d)106groupworkflowScan, diff, patch, and verify an MCP server stack across the 2025-11-25 to 2026-07-28 RC spec jump. Eight detection rules, AST rewrite via ts-morph, CLI-first. Foundation pillar by StudioMeyer (MCP Factory). TypeScript, MIT.
statusalphaupdated2026-05-30npm total136clones (14d)86groupfactorynpx -y mcp-stateless-migrator
Concepts
Repo-agnostic deep dives on the patterns behind the code.
- protocol
What is MCP, actually?
Model Context Protocol explained without the marketing. What it solves, what it doesn't, and when stdio beats HTTP.
Read → - architecture
Stdio vs HTTP for MCP
Transport choice changes everything. Latency, security model, deployment story. A practical decision tree.
Read → - memory
Memory architectures compared
Local SQLite + FTS5, knowledge graph, vector embeddings, hybrid. Which fits which agent shape?
Read → - protocol
Agent-to-Agent protocol v1.0 RC
What an agent-card.json buys you, what it doesn't, and how it relates to MCP and WebMCP.
Read → - browser
WebMCP for browser agents
W3C CG Draft, Chrome 146+. The proposal to give web pages a structured tool surface for in-browser agents.
Read → - agents
Agent orchestration patterns
Single agent, sequential pipeline, parallel fleet, judge-arbitrated. Trade-offs from running each in production.
Read →