@sunsetsyntax highlighted mcpls, a Rust MCP-to-LSP bridge so coding agents stop treating repos as raw text and inventing types. Agents get real hover, references, diagnostics, and rename via rust-analyzer, pyright, gopls, and more—local, private, and lightweight.
⚡ Key Takeaways
•Rust MCP-to-LSP bridge aimed at coding agents.
•Hooks rust-analyzer, pyright, gopls, and more for hover, refs, diagnostics, and rename.
•Runs locally and privately to cut hallucinated types from text-only repo views.
@DanKornas highlighted MARM Memory, a local-first MCP server so coding agents retain project context across chats and tools. It combines searchable local memory, repository indexing, and a concept graph; features include memory tools, a code graph, shared agent context, 14 tools over HTTP/STDIO, Pip/Docker deploy, Apache 2.0.
⚡ Key Takeaways
•Local-first MCP keeps decisions, research, fixes, and notes across sessions.
•Code graph indexes symbols and structure; concept graph links decisions, errors, and code.
•14 tools over HTTP/STDIO; Pip or Docker; Apache 2.0.
@Jolyne_AI highlighted Pulse, a lightweight edge sidebar that live-aggregates remaining quotas and rate-limit windows across 19 AI coding services including Claude Code, Codex, and Cursor, so you need not open each vendor page. Repo: https://github.com/qunqin24/Pulse.
⚡ Key Takeaways
•Edge sidebar aggregates remaining quotas across AI coding services.
•Covers about 19 tools including Claude Code, Codex, and Cursor, plus rate-limit windows.
•Open-source at qunqin24/Pulse to avoid checking each vendor site.
@aiRobertDaily highlighted fast-jev-compaction, a Claude Code plugin that replaces traditional LLM summarization with Jev-driven, decision-based context compaction. It quickly filters tool calls and results, drops or truncates stale content, and keeps only the necessary parts of valuable text; a related Codex-facing Jev context server was also noted. The aim is lower token cost and faster delivery.
⚡ Key Takeaways
•A Claude Code plugin replaces classic summarization with Jev decision-based compaction.
•Filters tool-call results, drops or truncates stale context, and keeps necessary signal.
•Targets lower token cost and faster delivery; a Codex-facing Jev path is also noted.
@DanKornas introduced Hypernovum, a desktop-only Obsidian plugin that maps project notes into a live 3D city using frontmatter, Git signals, and vault links. From one view you can launch Claude Code, Codex, Antigravity CLI, or a custom command, and inspect branch, commit velocity, dirty trees, stale projects, and merge conflicts. It is open source under AGPL-3.0.
⚡ Key Takeaways
•Project notes become a 3D city with status, priority, and backlink visual signals.
•Launch Claude Code, Codex, Antigravity CLI, or a custom command from a project.
•Git lens shows branch, dirty trees, stale work, and merge conflicts under AGPL-3.0.
@AverageAiBro highlighted slopscan by @jothantranston: a zero-dependency Node CLI that grades a GitHub repo or local folder for AI slop, sketchy agent-skill patterns, and basic hygiene, then prints a shareable S–F score card. Three bars cover AI Slop, Security, and Hygiene; it flags curl pipes in SKILL.md and credential grabs, can export a card image, and fail CI under a score you set. Scans stay local under MIT.
⚡ Key Takeaways
•Grade any GitHub repo or local folder with an S–F score via npx.
•Flags curl pipes in SKILL.md, injection phrases, and credential or webhook exfil patterns.
•Export a score-card image and gate CI with --fail-under; MIT and local-only scans.
@BennyKokMusic shared an open-source coding-agent harness that works with both Claude and Codex (plus OpenCode, Pi, and more) across web, desktop, and native iOS clients, runnable on a Mac or any cloud VPS. It adds agent-to-agent communication, image/video and file attachments, MCP-driven Codex/Claude sub-agents, and transcription for coding agents.
⚡ Key Takeaways
•One harness covers Claude, Codex, OpenCode, Pi, and more coding agents.
•Ships web, desktop, and native iOS clients for Mac or cloud VPS hosts.
•Adds MCP sub-agents, attachment display, and agent-to-agent communication.
@DanKornas introduced DispatchSEO, an open-source SEO workflow app for builders using Claude Code, Codex, or Cursor. It turns keyword ideas into content pull requests via agent research, an approval queue, scheduled builds, and search tracking. Agents connect over MCP to research, queues, rankings, pages, and Search Console; features include scheduled PR builds, daily rank checks, hourly Search Console snapshots, multi-site tokens, and an AGPL-3.0 license.
⚡ Key Takeaways
•MCP connects agents to research, queues, rankings, pages, and Search Console data.
•An approval queue feeds scheduled guide or free-tool pull-request builds.
•Multi-site deployments get separate tokens and data under AGPL-3.0.
@sunsetsyntax introduced mcp-gateway, a Rust binary aimed at MCP tool schemas that can burn about 15k tokens before an agent starts working. It exposes a fixed roughly 15-tool meta-surface, then discovers and invokes backends on demand for a local, private, low-bloat setup.
⚡ Key Takeaways
•A fixed ~15-tool meta-surface avoids dumping every MCP schema into context.
•Backends are discovered and invoked on demand in a local private process.
•Ships as a Rust binary focused on coding-agent token cost.
@DanKornas introduced agmsg, a local messaging tool for Claude Code, Codex, Gemini CLI, and GitHub Copilot CLI. Independent agent sessions coordinate through one shared SQLite store without a daemon, network service, or human copy-paste relay; durable history can be replayed into a fresh session, with monitor, turn, both, and manual delivery modes. The Bash + sqlite3 project is MIT-licensed and offers npx, plugin, and direct-script installation paths.
⚡ Key Takeaways
•A shared local SQLite transport lets independent CLI agent sessions hand work to one another directly.
•Durable history can be replayed into fresh sessions, with monitor, turn, both, and manual delivery modes.
•Bash + sqlite3, MIT open source, covering Claude Code, Codex, Gemini CLI, and Copilot CLI.
@AverageAiBro highlighted ondrej-merkun/skill-audit, a local scanner for prompt-injection and malware patterns across Claude Code, Codex, Copilot, Cursor, Gemini CLI, Windsurf, Cline, AGENTS.md, installed skills, plugins, and MCP configs. It ships 46 rules with no cloud upload, optional deeper review through localhost OpenAI-compatible models, CI support via --fail-on REVIEW, and a GitHub Action under the @ondrej-merkun/skill-audit npm package.
⚡ Key Takeaways
•Scans skills, plugins, MCP configs, and AGENTS.md locally across major coding-agent clients.
•Forty-six rules detect prompt injection and malware patterns without uploading configs to the cloud.
•Supports --fail-on REVIEW, a GitHub Action, and optional localhost model review for install and CI gates.
@iml1s covered xAI coding tool Grok Build adding /learn: it analyzes your agent chats, proposes new or updated Skills from repeated instructions, and flags stale rules, rarely used Skills, plugins, and MCP. Choose report-only or step-through apply. This is workflow-rule curation, not chat memory—akin to Claude Code /insights.
⚡ Key Takeaways
•/learn mines reusable Skills from chat so you stop retyping the same agent rules.
•Flags stale rules and cold Skills/plugins/MCP; report-only or step-through apply.
•Rule curation, not memory—Grok Build's take on Claude Code /insights.
@khyssabc highlighted iOfficeAI/OfficeCLI, an agent-first Office suite: one binary creates, reads, and edits DOCX/XLSX/PPTX with JSON output, batch/daemon modes, template merge, OpenXML validation, and HTML/PNG preview. It ships MCP for Claude Code and Codex, covers macOS/Linux/Windows via Homebrew/Scoop/npm, plus Python/Node SDKs, Apache 2.0. About 30k GitHub stars.
⚡ Key Takeaways
•Single binary reads/writes Word/Excel/PPT with no traditional Office install.
•MCP for Claude Code/Codex; OpenXML validation plus HTML/PNG preview.
•Apache 2.0; cross-platform binaries and Python/Node SDKs.
@DanKornas highlighted JKHeadley/instar, an open-source framework that pairs real Claude Code processes with persistent SQLite memory, cron jobs, and messaging so agents keep working across sessions. Features include SQLite/FTS5/rolling summaries with context re-injection, priority- and quota-aware scheduled work, Telegram/WhatsApp/iMessage, per-session tools/hooks/skills/MCP, plus behavioral hooks and external-operation gates. MIT.
⚡ Key Takeaways
•Real Claude Code processes plus SQLite memory so work survives across sessions.
•Cron with priority/quota awareness; Telegram/WhatsApp/iMessage outside the terminal.
•Per-session skills/MCP with safety gates; MIT open source.
@AverageAiBro covered Apple WebKit shipping a first-party Safari MCP server: any MCP client can attach to a Safari window for the DOM/a11y tree, network, console, screenshots, and about 17 page-interaction tools. Enable Allow remote automation and external agents, then point the client at /usr/bin/safaridriver --mcp (Claude, Codex, or mcp.json). Safari 27 beta brings the server beyond Technology Preview. Local only, no Apple network calls, and no AutoFill or personal browsing data.
@DanKornas highlighted TokenRhythm/opensquilla, a token-efficient microkernel AI agent for builders. SquillaRouter scores each turn locally on length, language, code, keywords, and semantic embeddings, then routes across four tiers to the cheapest capable model; the registry targets 20+ LLM backends with primary-plus-fallback. Fifteen bundled skills load on demand, and the agent can act as an MCP client or server. Local Markdown notes add SQLite full-text search and sqlite-vec semantic recall, with three sandbox policy tiers plus human-in-the-loop approvals for sensitive tools. Apache 2.0.
⚡ Key Takeaways
•Local four-tier routing picks the cheapest capable model instead of always using the top tier.
•Skills + MCP: 15 on-demand skills; agent can be MCP client and server.
•Apache 2.0; Markdown/SQLite memory with three sandbox tiers and human approvals.
@yibie covered Tencent open-sourcing Tencent/BrowserSkill: agents use your already-logged-in real browser inside a separate Agent Window—without stealing the main window—and must explicitly borrow open tabs then return them. It is harness-agnostic: Cursor, Claude Code, Codex, and other shell-capable clients go through bsk CLI → local daemon → browser extension. Humans can take over captchas and logins; 0.3.0 removed old bypass flags so extension settings own authorization.
⚡ Key Takeaways
•Real logged-in browser plus a separate Agent Window; tab borrow is explicit and returned.
•Harness-agnostic: shell clients drive the extension via bsk CLI and a local daemon.
•0.3.0 tightens auth: old confirm-bypass flags are gone; extension settings win.
@AverageAiBro covered GitLab 19.4 expanding the GitLab MCP server beyond lookups into a full loop: save pipelines, pull job logs, open/update/review/merge MRs, work items, commits, and Ultimate vulnerability tools. The same release extends one governance model—read-only defaults to Always allow; write and delete default to Always ask—for Duo agents and third-party MCP clients.
⚡ Key Takeaways
•MCP tools cover pipelines, job logs, full MR workflows, and vulnerability tools—not just lookups.
@DanKornas highlighted handmux/handmux, a self-hosted browser cockpit that drives the live tmux panes already on your machine from a phone or desktop browser instead of starting a separate remote session. It covers live history and input, streaming Codex chat (optional Claude Code/Pi), remote approval push notifications, Git and previews, plus QR session setup. AGPL-3.0.
⚡ Key Takeaways
•Exposes real local tmux panes so terminal and agent work continue away from the desk.
•Remote approvals: follow panes, get push notifications, and answer permission or plan prompts.
•Open-source handmux/handmux under AGPL-3.0; optional Codex, Claude Code, and Pi.
@DanKornas highlighted miracodeai/mira, a self-hosted AI code reviewer for teams that keep diffs, indexes, review history, and CVE data in their own SQLite or Postgres. It supports OpenRouter and OpenAI-compatible endpoints including vLLM/Ollama, applies per-repo rules via .mira.yaml, and auto-reviews PRs on GitHub, GitLab, and Forgejo/Codeberg. Apache-2.0.
⚡ Key Takeaways
•Self-hosted by default: code and review data stay on your infrastructure.
•Full-repo indexing plus custom rules go beyond a single diff.
•Supports GitHub, GitLab, and Forgejo; models via OpenRouter or local OpenAI-compatible endpoints.
