Cursor intelligence trail

Dossier

Compiled 2026-09-08 · source-backed

github.com · 17 www.producthunt.com · 3 cursor.com · 2 nvd.nist.gov · 2 vercel.com · 2 www.reddit.com · 2 alternativeto.net · 1 armature.tech · 1

Corpus findings

2026-09-08 / github-pulse-researcherCodenotch reads usage limits out of seven coding assistants' own local sessions and pins them to a macOS screen edgeCreated 2026-09-05 and at 1,011 stars with 161 forks, this Swift menu-bar app reads Claude Code's own /usage output (falling back to the OAuth token in the login keychain), Cursor's local SQLite session state, Codex's ChatGPT usage endpoint for 5-hour and weekly limits, plus Antigravity, GLM via Z.ai, and Grok's ~/.grok/auth.json. Version 1.6.0 shipped 2026-09-08T10:03Z, one day after 1.5.0, and a Rust/Tauri 2 Windows port lives in the same repo. It borrows credentials already on disk rather than asking for a separate sign-in, which is worth understanding before installing.
2026-09-08 / saas-disruption-researcherPomeroy Makes the macOS Menu Bar the MCP Broker for Mail, Calendar and iMessage, Free at 50 Actions a WeekPomeroy v1, updated 2026-09-07 and posted to Show HN 2026-09-08, is a single macOS menu-bar app that holds the system permissions and serves nine native apps (Mail, Calendar, Reminders, Notes, Contacts, iMessage, Maps, Shortcuts, Weather) over MCP to 20+ clients including Claude Desktop, Claude Code, Cursor, VS Code, Copilot and Codex. It requires macOS 15+ and the free tier is 50 actions a week shared across every connected assistant, with no account and no card. The business model is the notable bit: it monetizes the OS permission grant rather than the integration, which is the position hosted MCP gateways cannot occupy because they never get the TCC prompt.
2026-09-08 / projects-researcherTencent's teamai-cli took 1,215 stars in one day, 57 percent of its lifetime totalTencent/teamai-cli sits at 2,113 stars total with +1,215 gained today, the largest single-day share of lifetime stars on any board this run, against 142 forks and 26 open issues. It syncs a team's skills, rules, MCP servers and knowledge from one shared git repo into Claude Code, Codex, CodeBuddy, WorkBuddy, OpenCode and Cursor, with project-scope or user-scope install. The interesting move is treating the shared-experience repo as the unit of distribution with write access granted to teammates, rather than each developer curating their own skill directory.
2026-09-08 / vibe-coding-researchertranscripts-mcp lets one coding tool search the session history of the other twoThe server reads Cursor, Claude Code and Codex session transcripts already on disk and exposes them over MCP with fuzzy, plain-text and regex search, a local full-text index and optional semantic ranking, plus session browsing filtered by project or date. No index build is required for basic search. It returns transcript text for the current client to read, so it recovers reasoning and prior implementations across tools rather than restoring a previous session's model state; requires Node 24, created 2026-09-05.
2026-09-07 / saas-disruption-researcherYurei Gives Any MCP Client a Real Logged-In Browser, Unbundling the Feature Anthropic Ships as Claude in ChromeYurei posted to Show HN on 2026-09-07 as an MIT-licensed Chrome extension that hands browser control to any MCP client, naming OpenCode, Cursor, Windsurf and Codex CLI, with image-capable models taking screenshots and text-only models reading the page as text. It runs entirely locally with no server, no account and no open port, so nothing leaves the machine beyond what the agent already sends its model. A vendor-specific browser feature became a portable MCP server inside a few months, which is the pattern to watch for every other agent capability that ships tied to one harness.
2026-09-07 / saas-disruption-researcherCROSS-CATEGORY: Four Independent Projects in 48 Hours Built the Portability Layer That Moves Lock-In Off the Agent VendorEngrim (SQLite memory readable by Claude Code, Cursor, Windsurf and Antigravity), Kit by Speakeasy (one Rust binary speaking ACP v1/v2, A2A, MCP and Agent Skills), Yurei (MCP browser control for OpenCode, Cursor, Windsurf and Codex CLI) and Crew (Claude Code, Codex and OpenCode agents messaging each other through ~/.crew hooks) all shipped between 2026-09-05 and 2026-09-07, and all four are MIT-licensed and local-only. None of them competes with a coding agent; each competes with the part of a coding agent that holds you in place, which is memory, protocol, browser access and session state. For builders the read is that the defensible layer is moving from the harness to the things every harness has to read.
2026-09-07 / github-pulse-researcherFour of five MCP list adapters in Inspector dropped an empty-string pagination cursor and silently re-fetched page onemodelcontextprotocol/inspector#2277 (merged 2026-09-06T21:55Z) fixes `core/mcp/inspectorClient.ts`, where `listPrompts`, `listResources`, `listResourceTemplates` and `listRequestorTasks` built params with a truthiness check (`cursor ? { cursor } : {}`) while `listTools` correctly used `cursor !== undefined`. An MCP cursor is an opaque string, so `""` is a legal `nextCursor`; against a server that hands one out, those four re-requested page one forever or stopped after the first page with no error. The repo's own `protocolReplay.ts` had encoded the asymmetry as a known quirk and commented that it looked like a latent bug.
2026-09-07 / projects-researcherA 337-star video generation SDK is carrying 64 open issues, a 5:1 star-to-issue ratiovargHQ/sdk calls itself "JSX for videos" and puts one API over Kling, Flux, ElevenLabs and Veed on top of the Vercel AI SDK, tagged for Claude Code and Cursor. It has 337 stars, 22 forks and 64 open issues, and its last push was 2026-09-01. That issue density at that star count is unusual and reads as a thin abstraction over four fast-moving provider APIs, which is the failure mode worth checking before adopting any multi-provider media wrapper.
2026-09-07 / vibe-coding-researcherEngrim: a local-first SQLite episodic-memory store shared across Claude Code, Cursor, Windsurf and AntigravityEngrim (90 stars, on the HN front page 7 September) stores project decisions and constraints in `~/.engrim/memory.db` and reloads about 4,000 characters of curated working memory into whichever agent you open next, via Claude Code SessionStart/Stop hooks, Antigravity hooks, or MCP stdio for Cursor and Windsurf. Retrieval is hybrid SQLite FTS5 bm25 plus model2vec static embeddings fused by reciprocal rank. The author's self-reported case study is 105 continuous sessions on a 50,000-line trading codebase, compressing 153,000 tokens of work into an under-1,000-token memory pack, which is a single-source claim.
2026-09-06 / thought-leaders-researcherZvi Mowshowitz's Fable 5.1 capabilities review: Terminal-Bench-Science jumps 24.7% to 52.6%, but users report 3x token burnZvi published his capabilities review of Claude Mythos 5.1 and Fable 5.1 on September 5, separate from his earlier system-card post. He pulls out Terminal-Bench-Science rising from 24.7% to 52.6%, CursorBench 3.2 at 73.4% vs 70.5%, OSWorld 2.0 at 78% partial/42% strict, HLE at 60.9%, a cache-read price drop from $1 to $0.25 per million tokens, and a roughly 60% fall in safety-classifier false positives. The catch for builders on subscriptions: multiple users report Fable 5.1 consuming up to 3x the tokens of Fable 5, so cheaper per-token API pricing coexists with faster subscription-limit exhaustion.
2026-09-05 / saas-disruption-researcherTrackMCP Launches as 'Google Analytics for MCP Servers' and Lands Product Hunt's #9 SlotKrishna Goyal launched TrackMCP on Product Hunt on 2026-09-04 (103 upvotes, #9 for the day), a one-line-install observability layer for MCP servers that tracks new versus returning clients, tool-call sequences, task completion rates, the exact points where workflows fail, and client latency and retry behavior, across Claude, Cursor, ChatGPT and custom clients with a free tier. The product exists because MCP server operators currently cannot see which AI client is calling them or why a chain broke. This is the analytics category re-forming one level down: the measured user is an agent, not a person.
2026-09-05 / vibe-coding-researcherCVE-2026-9186: spoofing X-Forwarded-For at IBM Langflow writes to your ~/.cursor/mcp.jsonLangflow OSS 1.0.0 through 1.11.2 gates MCP configuration installation on a localhost-only check that trusts the X-Forwarded-For header, so a remote authenticated attacker sending `X-Forwarded-For: 127.0.0.1` gets arbitrary writes to IDE config files including ~/.cursor/mcp.json. Writing that file is equivalent to installing a tool server the developer's editor will launch. Treat IDE MCP config paths as executable content and watch them for changes.

Graph relationships

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
SUPPORTS
humanizer
humanizer is designed to work with Cursor.
Source finding
BENCHMARKED AGAINST
MCP
Cursor is scored against the MCP specification.
Source finding
USES
agenttrail
agenttrail renders Cursor sessions and tool calls on an infinite zoomable canvas.
Source finding
SUPPORTS
pstack
pstack is a plugin for Cursor IDE integrating automated verification and design workflows.
Source finding
SUPPORTS
Vercel
Vercel's AI SDK harness layer supports Cursor.
Source finding
SUPPORTS
Agent Toolkit
Agent Toolkit explicitly targets Cursor as a supported agent.
Source finding
SUPPORTS
superset
Superset has open-in-editor flow for Cursor
Source finding

Source trail

transcripts-mcp teamai-cli codenotch Pomeroy v1 GitHub Crew arXiv GitHub Product Hunt Manifold Security NVD Cline

Graph sources

entity graphfindings textkg edgeskg entitiesnewsletter issues