Commercial Agent Tools: October 19, 2026

1. Organ Runs a Company’s Departments as Scheduled Agent Teams

Organ. Organ, posted to Show HN on October 17, organizes agents into departments (strategy, product, R&D, marketing, operations) that run on schedules and produce plans, pull requests, and drafts, while the user approves merges, outbound email, and publishing. A dispatch router checks ownership before work proceeds, a security-officer role scopes each agent’s access to the repo, database, hosting, DNS, email, and API keys, and a shared “venture brain” turns agent observations into versioned memory, with issues escalated to an inbox by severity. It integrates with GitHub, GitLab, and Linear and uses Anthropic, OpenAI, or OpenRouter keys or Organ’s own metered key. Plans are $39, $99, and $299 per month with $15, $50, and $200 in included compute credits, and model tokens are billed separately. Source

2. DocuMemory Turns Coding Agent Sessions Into Searchable Memory Over MCP

DocuMemory. DocuMemory, posted to Show HN on October 18, is a hosted MCP server that distills a chat session into a markdown memory document (architecture decisions plus an ai_context_snippet for prompt injection) when the agent calls save_chat_as_doc. Retrieval combines semantic and full-text search with Reciprocal Rank Fusion on PostgreSQL with pgvector and HNSW indexes, and the company claims sub-50ms searches. It works with Claude Code, ChatGPT, Cursor, Windsurf, and any MCP client, exposes an OpenAPI spec, and offers a self-hosted option with the user’s own Postgres and a local model. Pricing is a 15-use free trial, then $12 per month or $120 per year, with team seats at $15 per month (two-seat minimum). Source

3. MeetCap Hits 1.0 as a Meeting Recorder That Agents Can Drive and Call

Seacen Zhao. MeetCap, released as v1.0.0 and posted to Show HN on October 18, is a macOS menu bar app and CLI that records meetings and runs speech recognition, live captions, and speaker diarization on the device. It installs an agent skill so Claude Code, Codex, or other agents can start and stop recordings, name speakers, fetch minutes, and search past meetings, and in the other direction it calls one of 20+ supported agent CLIs (including Codex, Cursor, Copilot CLI, and Qwen Code) to write live summaries and minutes on the user’s existing subscription, or uses an API key or a downloaded local model instead. Minutes link each point back to its transcript line and are saved as markdown. The app is free with no account under proprietary terms of use (the code is published on GitHub but not under an open-source license), and it requires macOS 14.2 or later. Source

4. Yila Packages Research Agents for Literature Review and Data Analysis

Yila AI. Yila, posted to Show HN on October 19 as “a Codex for scientific research,” offers task-specific research agents for academic work. Its literature agent searches several scholarly databases, claims to verify papers rather than invent citations, and reads 50+ full papers to draft citation-backed reviews, while its data agent profiles an uploaded CSV or XLSX, picks an analysis method, and returns charts plus reusable code. Other tools cover PDF Q&A, citation formatting in 2,800+ styles, figure, poster, and slide generation, and journal-format typesetting. The site does not name its underlying models, and it links to a paid pricing page (currently advertising a discount) without listing prices on the landing page. Source