OpenRig
Run teams of Claude Code, Codex and Pi agents
Agents & codingApache-2.0★ 8.6kv0.7.7
AI news
Perplexity can now use Graphify
Once the map exists, the questions get better.
GraphifyOfficial blogHow to use Graphify in Perplexity Computer
From the official blog (graphify.com), published . View original
GraphifyOfficial blogImpeccable v4.8.0
Impeccable skill v4.8.0 can be installed into Devin on the web, in the CLI and on desktop, and projects with a .replit file are identified as Replit Agent. Running impeccable detect --github in a GitHub Actions job annotates each finding on the line that caused it. Hard shadows offset by 4px or more are reported as advisories rather than failures. The launcher, the npm package and install compare each engine download against a digest list shipped with the skill and refuse a binary that does not match.
ImpeccableGitHub ReleasesAnythingLLM v1.19.0
AnythingLLM v1.19.0 can run a built-in local LLM inside Docker using AMD's open-source Lemonade inference server on llama.cpp, so Ollama or LM Studio are no longer required alongside it. The engine picks CUDA, Vulkan or the CPU automatically, runs on amd64 and arm64, and stays off until ENABLE_BUILT_IN_LLM=true is set. Color palettes can tint the whole interface. The release also caps request body size to block unauthenticated embed denial-of-service attacks.
AnythingLLMGitHub ReleasesBest AI Productivity Tools 2026: Assistants & Agents
From the official blog (affine.pro), published . View original
AFFiNEOfficial blogAFFiNE vs. AppFlowy vs. Anytype: The Ultimate Open-Source Comparison (2026 Edition)
From the official blog (affine.pro), published . View original
AFFiNEOfficial blog10 Free Weekly Schedule Templates for Any Week (2026)
From the official blog (affine.pro), published . View original
AFFiNEOfficial blogCline v0.2.0
Cline sdk v0.2.0 reattaches to a live session when one already has a run in progress instead of replacing its runtime, which previously left the original run going but unreachable. Model requests now retry at most 3 times after the first attempt, down from 5, and failed attempts are recorded in Langfuse traces. The gemini provider stops sending native Google Search, while Vertex keeps it. After the first compaction, the size estimate uses the messages as actually sent, fixing a roughly 7x overestimate that triggered compaction every few tool calls.
ClineGitHub Releases
Text-to-CAD
An open-source plugin that adds CAD modelling to coding agents and generates 3D models locally, MIT-licensed
Design & creativeMIT★ 24.9kv0.10.0
AnythingLLM
A local-first AI app that puts document chat, AI agents, scheduled jobs and a meeting assistant in one program; repository MIT-licensed, desktop installer under separate proprietary terms
Local AI & modelsMIT★ 67.8kv1.19.0
Rapid-MLX
An open-source LLM inference server for Apple Silicon built on MLX, with OpenAI- and Anthropic-compatible APIs and a focus on reliable tool calling for coding agents, Apache-2.0-licensed
Local AI & modelsApache-2.0★ 4.1kv1.0.1
Latitude
An open-source observability platform that traces AI agents in production and groups recurring failures into trackable signals, self-hostable and MIT-licensed
Dev tools & infraMIT★ 4.7kv7.18.0
Paperclip
An open-source, self-hosted app that runs AI agents as a company: org charts, budgets, goals and approvals, MIT-licensed
Agents & codingMIT★ 110.1kv2026.1023.0
Orca
An open-source agent IDE that runs Claude Code, Codex and other coding agents in parallel on the desktop, each in its own git worktree, on existing subscriptions; MIT-licensed
Agents & codingMIT★ 113.1kv1.4.246
T3 Code
An open-source control plane that runs Claude Code, Codex, Cursor and other coding agents from one interface on existing subscriptions; MIT-licensed
Agents & codingMIT★ 39.9kv0.1.6
OpenCode
An open-source coding agent for the terminal, desktop, web and IDEs, with 75+ model providers including local models; MIT-licensed
Agents & codingMIT★ 224kv2.2.5
Impeccable
An open-source skill and CLI that gives AI coding agents a design vocabulary: 24 design commands plus 62 deterministic detection rules, Apache-2.0 licensed
Design & creativeApache-2.0★ 100kv4.8.0
Hindsight
Agent memory that learns: retain, recall and reflect APIs, MIT-licensed and self-hostable
Dev tools & infraMIT★ 57.9kv1.1.0
Graphify
An open-source tool that turns a codebase with its docs, configs and PDFs into a knowledge graph AI coding assistants query through a skill or MCP; Apache-2.0, with pre-relicensing MIT contributions still under MIT
Dev tools & infraApache-2.0★ 140.8kv0.9.100
PageIndex
Vectorless, reasoning-based retrieval for long documents, with a tree index and page-level citations
Dev tools & infraMIT★ 40kv0.5.0
VoiceStudio
A local, open-source voice studio for cloning, dubbing, dictation, transcription and audiobooks; AGPL-3.0 app, non-commercial default OmniVoice weights
Local AI & modelsAGPL-3.0★ 98.5kv0.5.7
LM Studio
Find, download and run local LLMs with a desktop GUI
Local AI & modelsv0.4.28
Speech models in Ollama
Ollama can now turn text into speech.
OllamaOfficial blogMaking Buttondown easier to use
This time, we turned that lens on ourselves: the app you use to write, send, and manage your newsletter.
ButtondownOfficial blogPaperclip v2026.1023.0
Paperclip v2026.1023.0 lets a task or project be marked private and shared with named people and agents, so no one else can open it, find it in search, read its run transcripts or reach it through a plugin. Agents can run through your own model gateway, including OpenRouter, Amazon Bedrock, Google Gemini and endpoints compatible with OpenAI or Anthropic APIs, and Costs becomes an exact ledger that also lists monthly subscription fees. The Operator role preset grows from one grant to fourteen, including agent creation and configuration, so admins should review member roles. GitHub review bots move behind an experimental setting that is off by default, while existing bots keep running.
PaperclipGitHub ReleasesCoolify v4.5.0
Coolify v4.5.0 can deploy private Gitea repositories through Gitea App sources, introduces beta server patching with automatic security updates per server, and adds a fleet Metrics page with server and container tables and trend charts. Passkey sign-in, a file browser for applications, services and databases, and an automatic backup of the Coolify database before each update also arrive. Plaintext secrets are now encrypted at rest, while breaking changes require the deploy ability for API tokens that run scheduled tasks and similar operations and remove the POST /api/v1/enable endpoint.
CoolifyGitHub ReleasesClaude Code vs Codex (2026): Which Coding Agent to Use, and How to Run Them Together
From the official blog (lobehub.com), published . View original
LobeHubOfficial blogHindsight v1.1.0
Hindsight v1.1.0 introduces a single decision-model setup with per-operation overrides, adding Cloudflare Clef and the OpenAI Decisions API as providers, and recall ranks with that model when one is configured. An LLM request extension lets the server choose the model and credentials for each call. Document search covers document text and shows which fields matched. Webhook events gain delivery IDs, and fixes include the Windows export poll overrunning its timeout and unknown tags_match values being widened to any.
HindsightGitHub ReleasesWe Got Early Access to NVIDIA Nemotron 3.5 Embed. It's Hindsight's New Top Embedding Model.
Nemotron finished first on memory retrieval and first on the leaderboard, and it answered queries faster than a model just over half its size.
HindsightOfficial blog