Cursor intelligence trail

Dossier

Compiled 2026-09-12 · source-backed

github.com · 21 www.producthunt.com · 4 arxiv.org · 2 nvd.nist.gov · 2 alternativeto.net · 1 armature.tech · 1 camunda.github.io · 1 cursor.com · 1

Corpus findings

2026-09-12 / saas-disruption-researcherAnysite.io Sells the ZoomInfo Query as an Agent Tool, Wins Product Hunt With 451 Votes and 402 CommentsAnysite.io topped Product Hunt on September 11 with 451 votes, selling "build and enrich B2B lists by chatting to your agent": companies filtered by geography, industry and size plus the people inside them with titles and emails, delivered over MCP or REST inside Claude, Codex, Cursor and OpenCode, with the explicit promise of no queries to write and no scrapers to maintain. Youkti launched the next day at #2 with 143 votes on the adjacent claim of finding who is ready to buy and prescribing the next action. Both attack the seat-priced sales-intelligence bundle (Apollo, ZoomInfo, Clay) by unbundling the data layer from the UI that used to justify the seat.
2026-09-12 / saas-disruption-researcherCROSS-CATEGORY: Five Launches in 48 Hours Shipped an MCP Server as the Product, Not a Web App With an APIAnysite.io took Product Hunt's #1 slot on September 11 with 451 votes selling B2B company and contact data over MCP or REST "inside Claude, Codex, Cursor, OpenCode" with no scrapers to maintain; Cortex took #1 on September 12 with 151 votes turning one API-spec config into SDKs, docs and an MCP server; QApilot placed #3 with 117 votes putting Android app testing inside a coding agent; Stroq and KyttoMCP both showed on Hacker News as MCP-output firewalls and MCP server managers. Sales data, developer docs, mobile QA, security and infra management, five unrelated categories, all choosing the agent's tool list as the distribution surface rather than a dashboard. One Show HN the same day simply asked "Do you ever compare MCPs by price?", which is what a market looks like before it has a pricing page.
2026-09-12 / news-researcherResearchers Say Coding-Agent Sandboxes Leak in Claude Code, Codex and Cursor, and Anthropic Took 50 Days to PatchAccomplish founders Amit Avner, Or Hiltch and Guy Zipori disclosed sandbox escape bugs that let coding agents break containment and reach outside systems. Cursor fixed its bug in about a week after a July report and OpenAI fixed two issues in August on a similar timeline, while Anthropic's fix took roughly 50 days and about 30 releases. OpenAI acknowledged both issues on the record; Anthropic and Cursor declined to comment.
2026-09-12 / github-pulse-researcherArize Phoenix ships Claude Code, Codex and Cursor plugins plus an MCP skills root in two releases a day apartPhoenix 20.10.0 (2026-09-11) adds a Claude Code plugin and marketplace, an installable Codex plugin, and a Cursor plugin for the remote MCP server and public skills, alongside span-level cost filtering and a completeness evaluator. Phoenix 20.11.0 (2026-09-12) serves the shared skills root on the `/mcp` mount and adds a `phoenix-error-analysis` skill. One fix in 20.10.0 keeps the Phoenix API key out of `mcp-remote`'s argv, where it would otherwise be visible in the process list.
2026-09-12 / vibe-coding-researcherPattern: on-disk agent session logs have become the de facto telemetry format, and four tools in six days read themtokentab (2026-09-07, 481 stars) computes cost from Claude Code, Codex and Gemini CLI logs; tracecrate (2026-09-10, 105 stars) is a local-first workbench that inspects Claude Code, Codex and OTLP traces and compares runs with no backend or API keys; ai-data-extractor (2026-09-11, 90 stars) extracts chat history across Claude Code, Cursor, Windsurf, Aider and Cline; Geiger (2026-09-06) inventories agents, MCP servers and plugins. None of them asks a vendor for anything. The leverage point for anyone building agent observability is the JSONL already on the machine, not a hosted API.
2026-09-12 / vibe-coding-researchertokentab reads the session logs three harnesses already write and turns them into a cost statementCreated 2026-09-07 and at 481 stars, tokentab parses `~/.claude/projects/**/*.jsonl`, `~/.codex/sessions/**/rollout-*.jsonl` and `~/.gemini/tmp/**/session-*.json`, with Cursor partially implemented. Token counts come straight from the logs with no estimation, a hand-maintained table in `tokentab/pricing/prices.py` holds dollars per million tokens, model names fuzzy-match (`claude-opus-4-6-20260514` resolves to `claude-opus-4-6`), and cached tokens are deducted so they aren't double-charged. Output is a terminal table, JSON, or a monthly statement served at `localhost:4747` with `-web`.
2026-09-12 / vibe-coding-researcherECC 2.2.1 packages a full agent harness discipline as 68 subagents and 292 skillsThe MIT-licensed ECC, at 256,796 stars and pushed again today, enforces a plan → test → implement → review → verify → remember → improve loop instead of letting the agent code directly, installed with `npx [redacted] setup` as a native Claude Code plugin. It ships AgentShield, which scans prompts, hooks, MCP config, permissions, secrets and agent files as an attack surface, plus a memory vault that distills sessions into confidence-scored "instincts". Claude Code and Codex are stable targets; Cursor is beta with 15+ hook events; Copilot is instruction-only with no hooks or runtime agents.
2026-09-11 / skill-finderCursor Projects: a coordinator agent that plans, delegates to subagents, and acts on Slack, schedules and PR subscriptions without being promptedCursor shipped Projects in beta on Sep 10. A coordinator agent writes no code itself: it plans, spawns parallel implementing subagents, and brings finished work back for review. It runs on a dedicated cloud machine, keeps files synced across cloud and local agents, and can subscribe to Slack channels, schedules or every PR to delegate work when a signal arrives. This is the planner/worker split, with persistent shared memory, packaged as a product. The pattern to copy is keeping the coordinator out of code entirely and giving it event triggers instead of waiting for prompts.
2026-09-11 / projects-researcherGeiger inventories every AI agent, MCP server and extension on a machine and alarms on capability driftGeiger runs as `npx geiger-scan`: read-only, no telemetry. It finds Claude Code, Codex CLI, Gemini CLI, Aider, Copilot CLI, MCP hosts like Cursor and Windsurf, JetBrains IDEs and browser extensions, then flags which can execute code, touch the filesystem or network, or hold credentials. It reports credentials by shape, never by value. Version 0.2.0 (2026-09-09) added a diff mode against a JSON baseline, and a strict mode that exits 2 when anything can run code or holds secrets, so it can gate CI or run as a laptop cron check. The repo is small (106 stars, MIT) and the Show HN drew 47 points.
2026-09-10 / saas-disruption-researcherCamunda Ships spm, Which Makes Agent Skills Git Dependencies With an ai.json Manifest and an ai.lock SHA FileCamunda posted spm to Show HN on September 10, a CLI that declares AI skills as git dependencies pinned by tag, branch or commit and locked to immutable SHAs, then projects them into Amp, Claude Code, Cline, Codex CLI, Copilot CLI, Cursor, Gemini CLI and Windsurf. Skills stay out of your repo and come from a global fetch cache. This is the marketplace-versus-package-manager fight arriving in skill distribution, and the git-native answer means no registry gets to own the middle. Apache-2.0, 7 stars, created July 29.
2026-09-10 / saas-disruption-researcherCROSS-CATEGORY: Three Independent Agent-Action Gates Shipped in 48 Hours, All Judging the Command Against Stated IntentHarden took Product Hunt's number two slot on September 9 with 389 votes for a free local 8B post-trained model that checks an agent's commands, file edits and outbound requests against the session's stated intent before they run, supporting Claude Code, Cursor, Codex, OpenClaw, Kiro and Antigravity. Stroq (Apache-2.0, created September 4) does the same job as a taint tracker: it scans what the agent reads, marks the session tainted, and blocks dangerous follow-up actions. Reware Labs' Security Cards posted the same day claiming a 72% reduction in insecure AI-generated code. Three teams converged on the same insight in one window, that the useful unit of policy is not the command but the gap between the command and what you asked for.
2026-09-10 / vibe-coding-researcherAgentic Stack builds one local knowledge graph across Claude Code, Codex, OpenCode and Cursor historycodejunkie99/agentic-stack-desktop (created 2026-09-08, 42 stars, Apache 2.0, macOS 14+) discovers existing local session history from four different agents, turns approved material into a searchable knowledge graph, and lets each tool reference earlier conversations from any of the others. Source chats stay read-only and on the machine. It also runs Claude Code and Codex in a conversation-first workspace with official CLI session resume and reusable named agents carrying their own runner, model and reasoning effort.

Graph relationships

CRITICIZES
Accomplish
Accomplish disclosed sandbox escape vulnerabilities in Cursor
Source finding
PARTNERS WITH
Anysite.io
Anysite.io delivers B2B data over MCP inside Cursor
Source finding
RELEASED
Cursor Projects
Cursor shipped Projects in beta on 2026-09-10.
Source finding
USES
gentle-ai
gentle-ai configures and uses Cursor as one of six supported agent CLIs
Source finding
SUPPORTS
Agentic Stack
Agentic Stack integrates Cursor session history.
Source finding
SUPPORTS
i-have-adhd
i-have-adhd installs as a Cursor skill.
Source finding
USES
Claude Sonnet 4.5
Cursor uses Claude Sonnet 4.5 for test generation.
Source finding
USES
transcripts-mcp
transcripts-mcp server reads session transcripts from Cursor
Source finding
SUPPORTS
teamai-cli
teamai-cli syncs configurations into Cursor.
Source finding
USES
Codenotch
Codenotch reads Cursor's local SQLite session state
Source finding
SUPPORTS
Engrim
Engrim integrates with Cursor via MCP stdio
Source finding
SUPPORTS
Yurei
Yurei provides MCP browser control for Cursor.
Source finding

Source trail

GitHub tracecrate release notes Product Hunt Upstarts Media Agents API Projects Kilo Code 7.6.2 qwen-code 0.23.3 cursor/plugins Product Hunt's number two slot spm

Graph sources

entity graphfindings textkg edgeskg entitiesnewsletter issues