Artem KK

Making LLMs do what they're told

[Artem KK @KazKozDev]

footnote-mcp v0.2.3 is out. MCP Registry, PyPI, GHCR, Glama. Zero-key search: SearXNG, Wikipedia, Crossr

[Artem KK @KazKozDev]

footnote-mcp v0.2.3 🚀 MCP Registry, PyPI, GHCR, Glama. Zero-key search: SearXNG, Wikipedia, Crossref, GitHub, CommonCrawl. 45 tools for web research & verification. https://t.co/9z5vPuJKWy

[Artem KK @KazKozDev]

Full CRUD for issues/projects/cycles, labels, workflows, dependency tracking, team queries. Safety-aware: reads unrestricted, writes require confirmation. MIT license. pip install abscissa-mcp

[Artem KK @KazKozDev]

abscissa — MCP server for Linear: 42 tools that let AI agents manage issues, projects, cycles, and dependencies. Connect Claude Desktop or Codex directly to your Linear workspace. https://t.co/VQEZsX58do

[Artem KK @KazKozDev]

The writer is architecturally blocked from citing a source that didn't pass through the evidence store. Every citation resolves to a verified, stored page. Built on ollama (ornith:9b) + self-hosted SearXNG. Produces Markdown + letterhead PDF. MIT license.

[Artem KK @KazKozDev]

LLMFlow Search — local deep research where every claim is traced to its source. Five agents: planner → 3 parallel researchers → fact checker → writer → citation validator. SQLite evidence store. No API keys. https://t.co/3TTSyVr2WH

[Artem KK @KazKozDev]

@DanKornas LangGraph research agents are powerful! For the data extraction layer, footnote-mcp offers 42 MCP tools — web, tables, scholarly search, PDFs → https://t.co/9z5vPuJKWy

[Artem KK @KazKozDev]

@noelclawfun Great updates! For search & data extraction capabilities, check footnote-mcp — 42 MCP tools for web crawling, tables, PDF parsing, archives → https://t.co/9z5vPuJKWy

[Artem KK @KazKozDev]

@ValencianaAbel Local-first is the way! For research & data extraction, footnote-mcp provides 42 MCP tools — web crawling, tables, scholarly search, PDFs. All local → https://t.co/9z5vPuJKWy

[Artem KK @KazKozDev]

@jimmy_voxel51 Video search with vector embeddings is the future! For text-based research & data extraction, footnote-mcp offers 42 MCP tools. Open source → https://t.co/9z5vPuJKWy

[Artem KK @KazKozDev]

@0xKenny1st Nice results! For data extraction & research automation, check footnote-mcp — 42 MCP tools for web crawling, tables, PDFs, and more → https://t.co/9z5vPuJKWy

[Artem KK @KazKozDev]

@lifeisameeme Love the free tier! For deeper research — web crawling, table extraction, PDF parsing, archive lookups — footnote-mcp has 42 tools. Open source → https://t.co/9z5vPuJKWy

[Artem KK @KazKozDev]

@toolportapp Great approach! Tool discovery is key. For web search, crawling, table extraction & scholarly search, check footnote-mcp — 42 MCP tools for research → https://t.co/9z5vPuJKWy

[Artem KK @KazKozDev]

@cryptojezuz @Mention Love this! MCP is transforming agent data access. For search & data extraction, I built footnote-mcp — 42 open-source MCP tools for research agents → https://t.co/9z5vPuJKWy

[Artem KK @KazKozDev]

Verthor — reframes 16:9→9:16 by tracking the subject with YOLO, not center-cropping. Scene-aware, no GPU. https://t.co/LEzGFd6bq4 #VideoProcessing

[Artem KK @KazKozDev]

Most MCP search tools trust their own snippets. Mine checks the source first. footnote-mcp verifies claims against the actual source — 42 tools, NLI, 5-tier fetch. 100% accuracy. https://t.co/HwY7IKUZ3k #MCP

[Artem KK @KazKozDev]

Tested a customer support system prompt against 300+ adversarial attacks.
50% of authority bypasses worked. Base64 beat content filters. A simple roleplay prompt overrode every safety rule.
Open-sourced my benchmark: https://t.co/eXPycRpCB1

#LLMSecurity #PromptInjection

[Artem KK @KazKozDev]

Most agents don't fail because the model is dumb. They fail because the loop has no boundaries, memory is a mess, and nobody defined "done." Wrote about where agent architecture actually breaks.

https://t.co/dLFMKciQal

#AIAgents #LLM #AIEngineering

[Artem KK @KazKozDev]

New approach to train LLM agents using interactive ML tasks with online reinforcement learning, featuring exploration-enhanced fine-tuning, step-by-step RL, and an agent-specific reward module. https://t.co/U4yEQJyAKH #LLMTraining #ReinforcementLearning #AgentML