RAVENIntel
/api/ravenJSON

Research Index

21named reports • All sourced from raven-memory.md

Multilogin X Proxy Infrastructure Research
2026-03-09Infrastructure
DONE
  • •ISP proxies optimal: residential-grade auth at 60-70% lower cost than true static residential
  • •Datacenter proxies NON-STARTER for GMB (Google Aug 2025 Spam Update)
  • •Decodo (Smartproxy rebrand) shared ISP at $0.35/IP = $105/month for 300 IPs
  • •Webshare static residential at $0.30/IP = $90/month — cheapest viable option
  • +3 more findings
D:/ClaudeDev/00_GITHUB/_working-on/Tools/domain-portfolio-dashboard/.planning/research/multilogin-proxy-research.md
PBN Hosting Diversity Research
2026-02-28Link Building
DONE
  • •BulkBuyHosting best for true cPanel-per-account with unique IPs
  • •LaunchCDN best value at scale ($1.70-2.45/site via CDN approach)
  • •PBN.hosting best dashboard/features (Wayback restore, 400K+ IP pool)
  • •Priority Prospect most flexible (pay-per-IP, 46 locations/86 datacenters)
  • +3 more findings
D:/ClaudeDev/00_GITHUB/_working-on/Tools/pbn-hosting-diversity-intel-report.md
ChromaDB Vector Database Deep Dive
2026-03-06Vector DB
MINE FOR PATTERNS
  • •ChromaDB excels at prototyping but collapses under concurrent load (23s vs pgvector's 9.8s at 100 concurrent)
  • •HNSW index MUST fit in RAM — N = R * 0.245 (millions = GB RAM * 0.245)
  • •2025 Rust rewrite delivered 4x performance over Python-only version
  • •Hybrid search (BM25 + vector with RRF) is Chroma Cloud ONLY
  • +2 more findings
D:/ClaudeDev/00_GITHUB/_working-on/Tools/chromadb-deep-dive-intelligence-report.md
LanceDB Vector Database Technical Intelligence
2026-03-06Vector DB
ADOPT
  • •Strongest embedded vector DB for RAG — disk-first, no RAM ceiling
  • •Lance format: 100-1000x faster random access than Parquet
  • •Native hybrid search with 4 rerankers (RRF, Linear, Cohere, ColBERT) in open-source
  • •700M vectors in production at 77% cost reduction vs Milvus
  • +3 more findings
D:/ClaudeDev/00_GITHUB/_working-on/Tools/lancedb-technical-intelligence-report.md
Cloudflare AI Agent Infrastructure Stack
2026-02-25AI Discovery
DONE
  • •5 features: Markdown for Agents, Content-Signal Headers, LLMs.txt, AI Index, x402
  • •x402 processing 500K weekly transactions — not vaporware
  • •Cloudflare acquired Astro framework + Human Native (AI data marketplace) in Jan 2026
  • •Three competing payment protocols: x402 (Coinbase/CF), ACP (OpenAI/Stripe), AP2 (Google/Mastercard)
  • +2 more findings
D:/ClaudeDev/00_GITHUB/_working-on/Tools/cloudflare-ai-agent-research-brief.md
Cognee/Graphiti Evaluation — Memory Stack
2026-03-07Agent Memory
ADOPT
  • •Cognee and Graphiti solve DIFFERENT problems: document pipeline vs interaction memory
  • •Graphiti IS the engine behind Zep Cloud, open-sourced — full replacement
  • •Graphiti + Kuzu = zero-infrastructure temporal knowledge graph (no Neo4j needed)
  • •Kuzu: 18x faster ingestion than Neo4j, Cypher query language, works on Windows
  • +2 more findings
~/.claude/agents/souls/COGNEE-GRAPHITI-EVALUATION.md
OpenClaw DeepWiki Deep Dive
2026-03Infrastructure
DONE
  • •OpenClaw is a multi-channel AI agent GATEWAY (not a framework)
  • •4-layer agent execution pipeline with sophisticated failure handling
  • •Pi Agent SDK (@mariozechner/pi-agent-core) is the runtime — proprietary
  • •Dual memory backend: QMD (external) OR builtin (SQLite+FTS5+sqlite-vec, BM25+cosine hybrid, MMR, temporal decay)
  • +3 more findings
D:/ClaudeDev/00_GITHUB/_working-on/Tools/agent-soul-sync/research/openclaw-deepwiki.md
Agent Zero Deep Research
2026-03-12Infrastructure
ADOPT
  • •Viable VPS worker: HTTP + A2A + MCP triple protocol support
  • •SKILL.md compatible with our ecosystem
  • •AI-filtered memory system
C:/Users/mikem/.claude/agents/workspaces/raven/agent-zero-deep-research-2026-03-12.md
Paperclip Deep Research
2026-03-12Infrastructure
WATCH
  • •WATCH DON'T ADOPT assessment
  • •Steal: per-agent budget enforcement, audit log, heartbeat concept
C:/Users/mikem/.claude/agents/workspaces/raven/paperclip-deep-research-2026-03-12.md
CTRify Full Teardown
2026-03-26SEO Tools
SELECTIVE USE
  • •7.6M micro-worker CTR network (real residential IPs, 124 countries)
  • •NO GMB module — critical gap for local SEO
  • •$197-497/mo pricing
  • •AI content detectable, no API, 3-person team
  • +2 more findings
C:/Users/mikem/.claude/agents/workspaces/raven/ctrify-deep-research-2026-03-26.md
Outbound AI Calling Deep Research
2026-03-28Voice AI
ADOPT
  • •ElevenLabs native API: one POST request triggers call, no WebSocket server needed
  • •$0.11/min (EL $0.10 + Twilio $0.014)
  • •Custom LLM connects OpenClaw
  • •Batch calling API exists for scheduled/multi-call
  • +1 more findings
C:/Users/mikem/.claude/agents/workspaces/raven/outbound-ai-calling-deep-research-2026-03-28.md
Voice AI Pricing Comparison — 16 Platforms
2026-03-28Voice AI
DONE
  • •ElevenLabs: too expensive at $410/mo for 1-2hr/day sustained use
  • •Deepgram VA BYOM: $0.04/min — best managed option
  • •Hume Scale: $200/mo flat for 5K min — best subscription
  • •DIY Pipecat+Kokoro+Groq: $0.013/min — cheapest ($54/mo at 2hr/day)
  • +2 more findings
C:/Users/mikem/.claude/agents/workspaces/raven/voice-ai-pricing-comparison-2026-03-28.md
BYOK Voice AI Cost Breakdown
2026-03-28Voice AI
DONE
  • •SignalWire CONFIRMED no custom LLM (GPT-4o-mini/4.1-mini/4.1-nano only)
  • •Twilio phone layer: $0.018/min
  • •Best BYOK: $0.025/min (AssemblyAI+Kokoro+Groq)
  • •AssemblyAI STT: $0.0025/min (half of Deepgram)
  • +2 more findings
C:/Users/mikem/.claude/agents/workspaces/raven/byok-voice-ai-cost-breakdown-2026-03-28.md
Skill Eval Harness Market Opportunity
2026-03-28Market Intel
GO
  • •CONFIRMED GAP: zero domain-specific eval harnesses exist
  • •60K+ skills published, 7+ marketplaces, agencies building skills
  • •Anthropic skill-creator = generic. Tessl skill-optimizer = generic
  • •Security scanning crowded (5+ tools) but quality/domain testing = ZERO competitors
  • +2 more findings
C:/Users/mikem/.claude/agents/workspaces/raven/skill-eval-harness-market-opportunity-2026-03-28.md
Mem0 Multi-Agent Scoping Research
2026-03-29Agent Memory
DO NOT ADOPT
  • •AND queries across user_id+agent_id return EMPTY by design (#3218)
  • •90%+ implementations use user_id only
  • •Best pattern: metadata tags (user_id=human, metadata={agent:name})
  • •CrewAI drops agent_id silently
  • +1 more findings
C:/Users/mikem/.claude/agents/workspaces/raven/mem0-multi-agent-scoping-research-2026-03-29.md
Multi-Agent Memory Architecture Deep Research
2026-03-29Agent Memory
DONE
  • •18 tools analyzed, 6 papers, 40+ sources
  • •Letta: best native shared blocks (attach same block_id to N agents)
  • •Graphiti: best graph isolation (group_id per agent, bi-temporal tracking)
  • •Hindsight: 91.4% benchmark (bank_id), MIT, Docker — highest benchmark
  • +3 more findings
C:/Users/mikem/.claude/agents/workspaces/raven/multi-agent-memory-architecture-deep-research-2026-03-29.md
open-multi-agent Deep Research
2026-04-05Infrastructure
MINE FOR PATTERNS
  • •JackChen-me TypeScript multi-agent framework, 4844 stars in 5 days
  • •Coordinator pattern: LLM decomposes goal into task DAG
  • •4 stealable patterns: loop detection, coordinator prompt template, SharedMemory summary injection, retry token accumulation
  • •No MCP, no A2A, no persistence, single-process only
  • +1 more findings
C:/Users/mikem/.claude/agents/workspaces/raven/open-multi-agent-deep-research-2026-04-05.md
GeeLark Pricing + Proxy Intel
2026-04-03Infrastructure
DONE
  • •Free/Base($5)/Pro($19) tiers, 35% annual discount
  • •$0.007/min with $1/day cap, $29.90/mo rental for unlimited
  • •ADB works since v1.8.0 for text input
  • •Proxy recs: mobile > residential > datacenter for Reddit
  • +1 more findings
C:/Users/mikem/.claude/agents/workspaces/raven/geelark-pricing-proxy-intel-2026-04-03.md
tweakcc Deep Dive
2026-03-24Dev Tooling
ADOPT
  • •v4.0.11: ADOPT 3 features (MCP startup 50% faster, Opus Plan 1M, Session Memory auto-extract)
  • •Breaks on every CC update — reapply with npx tweakcc --apply
  • •1,413 stars, MIT license
C:/Users/mikem/.claude/agents/workspaces/raven/tweakcc-deep-dive-2026-03-24.md
DIYSmartCode Channel Research
2026-03-28Research
DONE
  • •120+ videos analyzed
  • •14 Tier 1 (must transcribe), 19 Tier 2, 22 changelogs
  • •Key topics: AutoDream, Subagents, Agent Teams, Skills, OpenClaw vs CC
C:/Users/mikem/.claude/agents/workspaces/raven/diysmartcode-channel-research-2026-03-28.md
Agentic Dashboard Research
2026-03-11Infrastructure
DONE
  • •5 dashboards + ecosystem analyzed
  • •Dominant stack 2026: Next.js + shadcn/ui + Tailwind + Recharts
  • •Mike owns $3-5K+ in template assets (33 Cruip, Tailwind Plus, 15 Aceternity, shadcn kit, 24 dashboards)
C:/Users/mikem/.claude/agents/workspaces/raven/agentic-dashboard-research-2026-03-11.md