Entity intelligence for any company, market, or question — not a chatbot that answers once, but a system that synthesizes with sources, turns each run into a reusable artifact, and watches for change later. Five-surface app + deep-work Workspace, Convex state, and a hosted public research MCP (npx nodebench-mcp).
Agent Claude Code
Full-stack deep diligence agent — structural QA, design coherence against target personas, narrative alignment, and competitive positioning audit. Covers code, UI, UX, content, performance, accessibility, and product-market fit in one pass.
Agent Claude Code
Full deployment, production verification, and launch readiness agent. Deploys Convex backend, Vercel frontend, tests voice server, runs production smoke tests, and produces a launch checklist.
Agent Claude Code
Self-dogfood NodeBench by using the deployed app and MCP tools, scoring quality, and filing findings.
Agent Claude Code
Runs benchmark suites, compares against baselines, and produces machine-readable and human-readable regression reports.
Agent Claude Code
Converts verified engineering and benchmark changes into draft changelog entries and release-note candidates.
Agent Claude Code
Produces structured strategic analysis for diligence, GTM, strategy, and intervention questions. Use when variables, scenarios, trust nodes, or ranked interventions matter.
Agent Claude Code
Tightens NodeBench UI for demo quality, executive clarity, and traversal reliability. Use for landing page, decision workbench, and operator-facing polishing.
Agent Claude Code
Supervises bounded continuous improvement loops for NodeBench. Use for roadmap-aligned iteration, benchmark review, and safe recurring maintenance.
Agent Claude Code
Full QA dogfood agent — traverses every surface, clicks every interactive element, checks contrast, performance, accessibility, behavioral correctness, and files findings as actionable P0/P1/P2 issues.
Agent Claude Code
Universal full-stack product diligence agent. Drop into any repo — audits UX, design, code quality, accessibility, performance, content, and competitive positioning. No app-specific knowledge needed.
Agent
This document outlines the complete restructure of the NodeBench AI agent system.
Agent
Before: The Morning Dossier page rendered raw RSS-style log lines ("Trending on Hacker News with 35 points and 6 comments") instead of editorial prose synthesis. Backend failures ("Pending — run this follow-up") leaked into the UI. No structured brief generation existed.
Agent
Make every in-app agent run capable of producing a final operator-facing verdict with open-source citations, trace-backed evidence, and explicit next actions, all surfaced through the existing UI.
Agent
A unified NodeBench subsystem enabling agents to self-heal from both semantic errors (rollback + lessons) and infrastructure failures (capability-aware model failover + budget gates), so long-running research jobs keep making progress when the user steps away.
Agent
This document converts the media-only material in Agentsetup/README.md into a maintainable operating spec and maps it onto the existing NodeBench harness.
Agent
Implemented comprehensive streaming UI optimization patterns to ensure smooth, animated streaming with per-step updates at 30-60fps without layout thrashing.
Agent
This document is the reference pattern for the deep-agent / agentic-system model in this repo.
Agent
This is the operating stance I want every coding agent to follow in this repo.
Agent
A complete streaming UI optimization system for FastAgentPanel that ensures smooth 30-60fps rendering with per-step updates and no layout thrashing.
Agent
This document describes the streaming UI optimization patterns implemented in FastAgentPanel to ensure smooth, animated streaming with per-step updates at 30-60fps without layout thrashing.