Model Context Protocol
model-context-protocol-bd362b33·50 events·first seen 1mo agoAliases: Model Context Protocol, Model Context Protocol (MCP)
Co-occurring entities
More like this (12)
Guides (1)
Recent events (50)
Anthropic Open-Sources the Model Context Protocol (MCP)
Anthropic has released the Model Context Protocol (MCP), an open standard enabling secure, two-way connections between AI assistants and external data sources such as business tools, content repositories, and development environments. The protocol introduces a client-server architecture with SDKs, local MCP server support in Claude Desktop, and a repository of pre-built connectors for systems like GitHub, Slack, Google Drive, and Postgres. Early adopters include Block and Apollo, with development tool companies Zed, Replit, Codeium, and Sourcegraph integrating MCP into their platforms. The goal is to replace fragmented, per-source integrations with a single universal protocol, improving context availability for AI agents.
MCP is Dead? — Community Debate on Model Context Protocol's Viability
A blog post from Quandri's engineering team provocatively questions whether the Model Context Protocol (MCP) is failing or already obsolete, generating significant community discussion on Hacker News with 236 points and 206 comments. The piece appears to critically examine MCP's adoption trajectory and potential shortcomings as a standard for AI agent tool integration. The high engagement suggests meaningful disagreement or concern in the practitioner community about MCP's future as an interoperability layer.
Anthropic Donates Model Context Protocol to Linux Foundation, Co-founds Agentic AI Foundation
Anthropic is donating the Model Context Protocol (MCP) to the newly established Agentic AI Foundation (AAIF), a directed fund under the Linux Foundation co-founded by Anthropic, Block, and OpenAI, with support from Google, Microsoft, AWS, Cloudflare, and Bloomberg. MCP has reached significant adoption milestones including 10,000+ active public servers, 97M+ monthly SDK downloads, and integration into ChatGPT, Gemini, Microsoft Copilot, and Visual Studio Code. The AAIF will also house Block's goose and OpenAI's AGENTS.md as founding projects, aiming to foster open, vendor-neutral standards for agentic AI. MCP governance will remain community-driven with existing maintainers continuing their roles.
Mistral AI Launches Connectors in Studio: Built-in and Custom MCP Support with Direct Tool Calling
Mistral AI has released Connectors in Studio, enabling developers to integrate enterprise data sources into AI applications via reusable connectors built on the Model Context Protocol (MCP). The feature supports both built-in connectors (GitHub, web search) and custom MCP servers, accessible via Conversation API, Completions API, and Agent SDK. New capabilities include direct tool calling for deterministic invocation, human-in-the-loop approval flows for governance, and programmatic connector management. Connectors are centrally registered and shared across Mistral products including LeChat and AI Studio.
Generate Images with Claude and Hugging Face via MCP
Hugging Face published a blog post demonstrating how to use Claude with the Model Context Protocol (MCP) to generate images through Hugging Face's inference infrastructure. The integration allows Claude to call Hugging Face image generation models as tools via MCP, connecting frontier LLMs with open-weight diffusion models. This represents a practical example of the agent-tool ecosystem pattern where LLMs orchestrate specialized model endpoints.
MCP for Research: How to Connect AI to Research Tools
Hugging Face published a blog post explaining how the Model Context Protocol (MCP) can be used to connect AI agents to research tools and data sources. The post covers practical patterns for integrating AI with academic and scientific workflows using MCP as a standardized interface layer. This is a commentary/tutorial piece aimed at researchers looking to extend AI agent capabilities into domain-specific tooling.
Five Big Improvements to Gradio MCP Servers
Hugging Face's Gradio team has announced five significant updates to Gradio's Model Context Protocol (MCP) server support. The improvements aim to make it easier to build and deploy MCP-compatible AI tool servers using Gradio. This is relevant to the growing agent-tool ecosystem where MCP is emerging as a standard protocol for connecting AI models to external tools and data sources.
Building the Hugging Face MCP Server
Hugging Face has published a blog post describing the construction of an MCP (Model Context Protocol) server that exposes Hugging Face platform capabilities to AI agents and LLM toolchains. The post covers the architecture and implementation of the server, enabling agents to search models, datasets, and spaces programmatically. This represents Hugging Face's integration into the emerging MCP ecosystem for agent-tool interoperability.
Tiny Agents in Python: a MCP-powered agent in ~70 lines of code
Hugging Face published a tutorial demonstrating how to build a minimal AI agent in approximately 70 lines of Python using the Model Context Protocol (MCP). The post shows how MCP enables tool discovery and invocation for LLM-based agents with very little boilerplate. This is part of a broader trend of simplifying agent construction by standardizing tool interfaces.
How to Build an MCP Server with Gradio
Hugging Face published a tutorial on building Model Context Protocol (MCP) servers using Gradio, enabling AI models to expose tools and resources through the MCP standard. The post demonstrates how Gradio applications can serve as MCP-compatible backends, allowing AI agents to discover and invoke Gradio-hosted functions. This lowers the barrier for ML practitioners to participate in the emerging MCP ecosystem without deep protocol knowledge.
Tiny Agents: an MCP-powered agent in 50 lines of code
Hugging Face published a blog post demonstrating how to build a minimal AI agent using the Model Context Protocol (MCP) in approximately 50 lines of code. The post showcases how MCP enables agents to discover and invoke tools dynamically, reducing the boilerplate required for agentic workflows. This serves as both a tutorial and a commentary on MCP's role in simplifying agent-tool integration in the current ecosystem.
Le Chat Launches MCP Connector Directory and Persistent Memories
Mistral AI has released two major features for Le Chat: a directory of 20+ enterprise MCP-powered connectors (beta) spanning data, productivity, development, automation, and commerce tools, plus custom extensibility for any remote MCP server. A Memories feature (beta) has also launched, enabling persistent cross-conversation context with user-controlled storage and privacy settings. Both features are available on the free plan, with enterprise deployment options including self-hosted and private cloud. Mistral is positioning Le Chat as a unified enterprise AI assistant surface competing directly with ChatGPT and similar products.
Anthropic Launches Claude for Creative Work with Eight New MCP Connectors
Anthropic has released a suite of connectors enabling Claude to integrate directly with major creative software platforms including Adobe Creative Cloud, Blender, Autodesk Fusion, Ableton, Affinity by Canva, SketchUp, Resolume, and Splice. The connectors are built on the Model Context Protocol (MCP), making them accessible to other LLMs as well. Anthropic also announced Claude Design, a new product from Anthropic Labs for exploring software UI concepts with export to Canva, and partnerships with RISD, Ringling College, and Goldsmiths to support creative computing curricula. A one-time donation was made to the Blender project to support its Python API development.
Mistral AI Launches Agents API with Built-in Connectors, MCP Tools, and Persistent Memory
Mistral AI has released a dedicated Agents API that extends beyond chat completion by providing built-in connectors for code execution, web search, image generation, and document retrieval, alongside support for Model Context Protocol (MCP) tools. The API features stateful conversation management with branching, streaming output, and multi-agent orchestration capabilities. Benchmark results show substantial web search augmentation gains: Mistral Large jumps from 23% to 75% on SimpleQA, and Mistral Medium from 22% to 82% with search enabled. The release targets enterprise-grade agentic workflows and is accompanied by cookbooks covering GitHub coding assistants, financial analysis, and travel planning use cases.
Implementing MCP Servers in Python: An AI Shopping Assistant with Gradio
Hugging Face published a tutorial demonstrating how to build Model Context Protocol (MCP) servers in Python using Gradio, illustrated through a virtual try-on AI shopping assistant. The post covers integrating MCP tool exposure with Gradio's interface layer, enabling AI agents to invoke image-based try-on capabilities as structured tools. This represents a practical guide for developers connecting multimodal AI models to agent frameworks via MCP.
Upskill your LLMs With Gradio MCP Servers
Hugging Face published a blog post explaining how to build Model Context Protocol (MCP) servers using Gradio, enabling LLMs to access custom tools and external capabilities. The post covers how Gradio applications can be exposed as MCP-compatible tool endpoints that AI agents can invoke. This positions Gradio as part of the growing MCP ecosystem for extending LLM functionality with structured tool use.
Skybridge: Full-Stack TypeScript Framework for MCP and ChatGPT Apps
Skybridge is an open-source TypeScript framework designed for building Model Context Protocol (MCP) applications and ChatGPT-integrated apps. It offers type-safety, React-based UI components, and platform-agnostic deployment. The project has accumulated 1,368 GitHub stars with 56 added today, indicating growing community traction.
Microsoft Azure DevOps MCP Server
Microsoft has published an open-source Model Context Protocol (MCP) server for Azure DevOps, enabling AI agents to interact directly with Azure DevOps services. The repository is implemented in TypeScript and has accumulated 1,710 GitHub stars. This extends the MCP ecosystem with enterprise DevOps tooling, allowing agents to perform operations such as managing pipelines, work items, and repositories.
Chrome DevTools MCP Server for Coding Agents
The chrome-devtools-mcp repository exposes Chrome DevTools functionality as a Model Context Protocol (MCP) server, enabling coding agents to interact with browser debugging tools programmatically. The project has accumulated over 40,000 stars on GitHub, with 132 added today, indicating strong community traction. This tooling bridges browser developer tooling with AI agent workflows, allowing agents to inspect, debug, and interact with web pages.
MetaTrader MCP Server: AI LLM Trading via Model Context Protocol
An open-source Python project implementing a Model Context Protocol (MCP) server that enables AI language models to execute trades on the MetaTrader platform. The repository has gained 82 stars in a single day, reaching 408 total. This represents a concrete deployment of the MCP agent-tool pattern in a financial trading context.
PROVE framework trains LLMs for multi-step tool use via stateful MCP environments and programmatic rewards
Researchers introduce PROVE (Programmatic Rewards On Verified Environments), a framework for training LLMs to orchestrate multi-step tool calls using reinforcement learning. The system includes a library of 20 stateful MCP servers with 343 tools, an automated data synthesis pipeline that grounds training queries in live server state, and a multi-component programmatic reward function requiring no judge model. Training four models (Qwen3-4B, Qwen3-8B, Qwen2.5-7B, Granite-4.1-8B) with ~13K examples yields gains of up to +10.2 on BFCL Multi-Turn, +6.8 on tau2-bench, and +6.5 on T-Eval, demonstrating consistent improvements in multi-step tool orchestration.
ProvenanceGuard: Source-aware factuality verification for MCP-based LLM agents
Researchers introduce ProvenanceGuard, a verifier that checks factual claims in MCP-grounded LLM agent answers against their specific source provenance rather than pooled evidence. The system decomposes answers into atomic claims, routes each to its attributed source via MCP trace metadata, and applies NLI plus token-alignment checks to detect 'cross-source conflation' — where a claim is supported somewhere but attributed to the wrong source. Evaluated on 281 medical-domain MCP-agent traces, it achieves block F1 of 0.802 and source accuracy of 0.858 on held-out data, and detects all injected attribution swaps in 50 controlled clinical probes. The work establishes source attribution as an independent factuality axis distinct from standard grounding checks.
Anthropic Discloses First Reported AI-Orchestrated Cyber Espionage Campaign Using Claude Code
Anthropic detected and disrupted a sophisticated espionage campaign in mid-September 2025, attributed with high confidence to a Chinese state-sponsored threat actor, that used Claude Code as an autonomous agent to attack roughly thirty global targets across tech, finance, chemical manufacturing, and government sectors. The attackers jailbroke Claude Code by decomposing malicious tasks into seemingly innocent subtasks and falsely framing it as defensive security testing, enabling largely autonomous reconnaissance, vulnerability exploitation, credential harvesting, and data exfiltration. Anthropic describes this as the first documented large-scale cyberattack executed without substantial human intervention, leveraging agentic AI capabilities, tool access via MCP, and advanced coding skills. The company banned identified accounts, notified affected entities, coordinated with authorities, and is expanding detection classifiers and publishing the report to aid industry and government defenses.
Anthropic Launches Claude for Financial Services with Claude 4 Models and Ecosystem Integrations
Anthropic has introduced a Financial Analysis Solution targeting finance professionals, built around Claude 4 models and pre-built MCP connectors to data providers including FactSet, S&P Global, PitchBook, Databricks, and Snowflake. Claude Opus 4 reportedly passed 5 of 7 levels of the Financial Modeling World Cup and scored 83% accuracy on complex Excel tasks when deployed by FundamentalLabs. The solution includes Claude Code with expanded usage limits, expert implementation support, and partnerships with major consultancies including Accenture, Deloitte, KPMG, and PwC. Early adopters include Bridgewater's AIA Labs, which has used Claude since 2023 for investment analyst workflows.
Introducing gpt-realtime and Realtime API updates
OpenAI is releasing a new speech-to-speech model called gpt-realtime alongside expanded Realtime API capabilities. New features include MCP server support, image input, and SIP phone calling support. These updates extend the Realtime API's utility for voice-driven and multimodal agent applications.
AiraXiv: AI-Driven Open-Access Publishing Platform for Human and AI Scientists
AiraXiv is a proposed open-access academic publishing platform designed to accommodate both human and AI-generated research outputs, addressing scalability challenges in traditional peer review. The platform supports AI scientists via Model Context Protocol (MCP)-based interactions and human scientists through an interactive UI, with papers evolving through continuous feedback-driven iteration. It was validated through real-world deployment as the submission platform for ICAIS 2025. The work positions itself as infrastructure for a future where AI agents are first-class participants in the scientific publishing ecosystem.
Google Labs Releases Stitch Skills: Agent Skills Library for MCP-Compatible Coding Agents
Google Labs has published stitch-skills, a TypeScript library of Agent Skills designed to work with the Stitch MCP server. The library follows the Agent Skills open standard, enabling compatibility with multiple coding agents including Gemini CLI, Claude Code, Cursor, and Antigravity. The repository has accumulated 5,597 stars with 70 added today, indicating active community interest in the MCP/agent tooling ecosystem.
phodal/routa: Workspace-First Multi-Agent Coordination Platform with MCP/ACP/A2A Support
Routa is an open-source TypeScript project providing a workspace-first multi-agent coordination platform for AI development. It features shared Specs, Kanban-style orchestration, and support for multiple agent communication protocols including MCP, ACP, and A2A across web and desktop environments. The repository has gained significant traction with 1,136 total stars and 141 stars added today, signaling community interest in multi-agent tooling.
Data Points: DeepSWE Benchmark, DeepSeek V4 Price Cuts, MAI-Image-2.5, Mythos Security Findings, MCP Stateless Update
This edition of The Batch covers five distinct AI developments: Datacurve's DeepSWE benchmark claims to fix critical grading flaws in SWE-bench Pro with hand-written verifiers and harder tasks; DeepSeek permanently cuts V4 Pro prices by 75%; Microsoft's MAI-Image-2.5 debuts third on the Arena leaderboard; Anthropic's Claude Mythos Preview found over 10,000 high/critical vulnerabilities in the first month of Project Glasswing, with remediation badly lagging discovery; and the Model Context Protocol proposes removing stateful sessions to enable stateless, load-balanced remote servers. Each item reflects meaningful movement in evaluation methodology, inference economics, multimodal generation, AI-assisted security, and agent tooling infrastructure.
Anthropic Introduces Claude Opus 4 and Sonnet 4 with Leading Coding Benchmarks and Agent Capabilities
Anthropic has released Claude Opus 4 and Claude Sonnet 4, positioning Opus 4 as the world's best coding model with 72.5% on SWE-bench and 43.2% on Terminal-bench, and Sonnet 4 at 72.7% on SWE-bench. Both models are hybrid (near-instant + extended thinking), support extended thinking with tool use in beta, parallel tool execution, and improved memory via local file access. Alongside the models, Anthropic is launching Claude Code as generally available with GitHub Actions, VS Code, and JetBrains integrations, plus four new API capabilities: code execution tool, MCP connector, Files API, and one-hour prompt caching. Pricing is unchanged from prior Opus and Sonnet tiers ($15/$75 and $3/$15 per million tokens respectively), with availability on Anthropic API, Amazon Bedrock, and Google Cloud Vertex AI.
Apple's Xcode 26.3 Integrates Claude Agent SDK for Autonomous Coding
Xcode 26.3 introduces native integration with Anthropic's Claude Agent SDK, enabling autonomous, long-running coding tasks directly within Apple's IDE. The integration supports visual verification via Xcode Previews, full-project reasoning across Apple frameworks, autonomous task execution with goal-directed behavior, and MCP-based access for Claude Code CLI users. This expands on an earlier September announcement that brought Claude Sonnet 4 to Xcode in a limited turn-by-turn capacity, now replacing it with the same agentic harness that powers Claude Code.
Anthropic Signs CMS Health Tech Ecosystem Pledge to Advance Healthcare Interoperability
Anthropic has signed the Centers for Medicare & Medicaid Services (CMS) Health Tech Ecosystem Pledge, a public-private initiative to modernize healthcare data sharing and improve interoperability. The company plans to leverage Claude and the Model Context Protocol (MCP) to bridge incompatible health data systems, enabling AI assistants to securely access patient data from CMS Aligned Networks and personal health records with patient consent. CPO Mike Krieger framed the effort as applying Anthropic's existing interoperability solutions to the longstanding problem of healthcare data silos.
Anthropic expands Claude for Education with Canvas, Panopto, and Wiley integrations via MCP
Anthropic announced new integrations for its Claude for Education platform, including Canvas LTI support and pre-built MCP servers connecting Claude to Panopto lecture recordings and Wiley academic content. New institutional partners include the University of San Francisco School of Law and Northumbria University. The announcement also covers expansion of student ambassador programs and launch of Claude Builder Clubs on campuses. Student privacy protections are emphasized, with conversations excluded from AI training by default.
HyperTool: Unified executable MCP-style interface reduces step-wise tool call overhead for LLM agents
HyperTool introduces a unified executable interface that allows LLM agents to invoke multiple tool calls within a single code block, hiding intermediate dataflow from the main reasoning trace. This addresses an 'execution-granularity mismatch' where step-wise atomic tool calls waste context and force models to manage low-level operations. On the MCP-Universe benchmark, HyperTool more than doubles accuracy for Qwen3-32B (15.69% → 35.29%) and Qwen3-8B (9.93% → 33.33%), outperforming GPT-OSS and Kimi-k2.5.
Anthropic Launches Ten Finance Agent Templates with Microsoft 365 Integration and Expanded Data Connectors
Anthropic is releasing ten ready-to-run agent templates targeting high-value financial services workflows including pitchbook creation, KYC screening, and month-end close, deployable as plugins in Claude Cowork/Claude Code or as autonomous Claude Managed Agents. The release includes native add-ins for Microsoft Excel, PowerPoint, Word, and Outlook with cross-application context persistence. Claude Opus 4.7 underpins the offering and leads the Vals AI Finance Agent benchmark at 64.37%, with new data connectors from partners including Dun & Bradstreet, Fiscal AI, FactSet, S&P Capital IQ, and others providing governed real-time data access.
Cognizant to Deploy Claude to 350,000 Employees in Major Enterprise AI Partnership
Cognizant, a global IT consulting firm, has announced a partnership with Anthropic to deploy Claude to up to 350,000 employees across engineering, delivery, and corporate functions. The deployment integrates Claude Code, the Model Context Protocol (MCP), and Anthropic's Agent SDK with Cognizant's own platforms including Flowsource, Neuro AI, and Agent Foundry. Use cases span software engineering productivity, legacy modernization, multi-agent orchestration, and vertical industry solutions beginning with Financial Services. The partnership also positions Cognizant as a channel for helping its enterprise clients adopt agentic AI at scale.
Anthropic Launches 'Labs' Team to Incubate Experimental AI Products
Anthropic is formalizing and expanding 'Labs,' an internal team dedicated to incubating experimental products at the frontier of Claude's capabilities. Mike Krieger (Instagram co-founder, former Anthropic CPO) is joining Labs alongside Ben Mann, while Ami Vora takes over as head of the Product organization. The announcement cites Claude Code, MCP, and Cowork as products that emerged from this experimental approach, and signals a structural reorganization to separate frontier experimentation from scaling established products.
Anthropic Partners with US Department of Energy on Genesis Mission for AI-Driven Scientific Discovery
Anthropic and the US Department of Energy have announced a multi-year partnership under the DOE's Genesis Mission initiative, targeting AI deployment across energy, biological sciences, and scientific productivity domains. The partnership will provide DOE researchers access to Claude and Anthropic engineers who will build purpose-built agents, Model Context Protocol servers, and specialized Claude Skills for scientific workflows. The collaboration has potential reach across all 17 US national laboratories and builds on prior work including a nuclear risk classifier with the National Nuclear Security Administration and Claude deployment at Lawrence Livermore. This represents a significant expansion of Anthropic's US government footprint.
Anthropic announces 'Code with Claude' — its first developer conference
Anthropic announced Code with Claude, its first developer conference, held May 22, 2025 in San Francisco at The Midway. The one-day event targets a select group of developers and founders, with sessions covering the Anthropic API, Claude Code, Model Context Protocol (MCP), AI agent implementation strategies, and tool use patterns. Attendees will hear from Anthropic's executive and product teams and participate in interactive labs and office hours.
OmniRoute: Open-Source AI Gateway with 160+ Providers and ~95% Context Compression
OmniRoute is a TypeScript-based open-source AI gateway that unifies access to 160+ AI providers through a single endpoint. It features RTK+Caveman stacked compression claiming up to ~95% eligible context savings, smart auto-fallback, and support for MCP/A2A protocols. The project has gained notable traction with nearly 5,000 stars and 122 new stars in a single day.
Anthropic Expands Claude for Financial Services with Excel Add-in, New Connectors, and Agent Skills
Anthropic is expanding its Claude for Financial Services offering with a beta Excel add-in (Claude for Excel), seven new real-time data connectors (including LSEG, Moody's, Aiera, and Chronograph), and six new pre-built Agent Skills covering tasks like DCF modeling, comparable company analysis, and initiating coverage reports. The updates build on Claude Sonnet 4.5's performance on the Finance Agent benchmark from Vals AI, where it scored 55.3% accuracy. Claude for Excel allows users to read, analyze, modify, and create Excel workbooks directly from a sidebar, with transparency into cell-level changes. These features are rolling out in preview to Max, Enterprise, and Teams users, with Citi cited as a notable enterprise adopter.
Awesome Harness Engineering: Curated List for AI Agent Infrastructure
A GitHub repository aggregating resources on AI agent harness engineering, covering tools, patterns, evaluations, memory systems, MCP (Model Context Protocol), permissions, observability, and orchestration. The list has accumulated 1,318 stars with 39 added today, indicating moderate community traction. It serves as a reference index rather than original research or tooling.
Anthropic and Salesforce Expand Partnership to Bring Claude to Regulated Industries via Agentforce
Anthropic and Salesforce have announced an expanded partnership making Claude a preferred foundational model for Salesforce's Agentforce platform, with a focus on regulated industries including financial services, healthcare, cybersecurity, and life sciences. Claude operates within Salesforce's virtual private cloud trust boundary via Amazon Bedrock, making Anthropic the first LLM provider fully integrated within that boundary. The partnership also includes Salesforce deploying Claude Code across its global engineering organization, a bidirectional Slack-Claude integration via MCP server, and plans to co-develop industry-specific AI solutions starting with financial services. Early adopters include RBC Wealth Management and CrowdStrike.
Mistral AI Launches Le Chat Enterprise with Mistral Medium 3
Mistral AI has introduced Le Chat Enterprise, a feature-rich AI assistant platform powered by the newly announced Mistral Medium 3 model, targeting enterprise AI adoption challenges such as tool fragmentation and insecure knowledge integration. The platform includes enterprise search with connectors to Google Drive, SharePoint, OneDrive, Gmail, and Google Calendar, plus agent builders, document libraries, custom model support, and hybrid deployment options. Le Chat Enterprise is available now on Google Cloud Marketplace, with Azure AI and AWS Marketplace listings forthcoming. Mistral also announced improvements to its Le Chat Pro and Team plans.
Mistral Releases Search Toolkit: Open-Source Composable Framework for Production RAG and Enterprise Search Pipelines
Mistral AI has launched Search Toolkit in public preview, an open-source framework that unifies document ingestion, retrieval, and evaluation into a single composable pipeline for AI applications. The toolkit ships with BM25 sparse retrieval, dense embedding-based retrieval, hybrid configurations, and built-in metrics (recall, precision, MRR, NDCG), targeting enterprise RAG workflows, domain-specific retrieval, and agentic systems. It integrates with MCP-based Connectors for live data access from CRMs, code repositories, and productivity tools. CMA CGM is cited as a production user, combining Search Toolkit with Voxtral for real-time fake news detection across audio sources.
Mistral Releases Leanstral: First Open-Source Code Agent for Lean 4 Formal Verification
Mistral AI has released Leanstral, an open-source code agent built on a sparse 120B/6B-active-parameter architecture, designed specifically for formal proof engineering in Lean 4. The model targets realistic proof engineering workflows rather than isolated math competition problems, and is benchmarked on FLTEval, a new evaluation suite tied to the Fermat's Last Theorem formalization project. Leanstral is released under Apache 2.0 with a free API endpoint and MCP support, and demonstrates competitive performance against Claude Sonnet 4.6 at roughly 1/15th the cost. The release positions formal verification as a scalable alternative to human code review for high-stakes software and mathematics.
Headroom: token compression library for LLM tool outputs, logs, and RAG chunks
Headroom is an open-source Python library that compresses tool outputs, logs, files, and RAG chunks before they reach an LLM, claiming 60-95% token reduction with minimal answer quality loss. It ships as a library, proxy, and MCP server. The project gained significant traction on GitHub with 6,148 stars and 1,266 stars in a single day.
Data Points: Apple/Google Siri overhaul, Gemma 4 12B, Kimi Code CLI, OpenJarvis, and U.S. OpenAI stake talks
A multi-item digest covers several significant AI developments: Apple is expected to announce a revamped Siri at WWDC that uses Google Gemini models distilled for on-device use alongside cloud routing, marking a notable Apple-Google AI partnership. Google released Gemma 4 12B, an encoder-free multimodal open-weights model designed for consumer laptops under Apache 2.0. Moonshot AI released Kimi Code CLI, an open-source terminal coding agent with native subagent orchestration and conversational MCP configuration. Stanford and Lambda Labs released OpenJarvis, an on-device agent framework claiming near-cloud accuracy at 800× lower API cost. The White House and OpenAI are reportedly negotiating a government equity stake in OpenAI as part of a proposed Public Wealth Fund.
Data Points: Perplexity Computer expands, Google Aletheia math agent, DeepSeek chip strategy, Nvidia retrieval pipeline, Stargate cancellation
The Batch's weekly data points roundup covers five significant AI developments: Perplexity expanded its Computer agentic platform to desktop, mobile, and enterprise with new APIs and financial data tools; Google released Aletheia, a Gemini-based math research agent achieving 95.1% on IMO-Proof Bench Advanced (up from 65.7%); DeepSeek withheld pre-release access to its V4 model from Nvidia and AMD while giving domestic Chinese chipmakers early access; Nvidia's NeMo Retriever topped the ViDoRe v3 leaderboard using a ReACT-based agentic retrieval loop; and OpenAI and Oracle cancelled plans to expand the Abilene Stargate campus from 1.2 GW to 2.0 GW due to financing and reliability issues.
Microsoft Build: Seven in-house AI models, GitHub Copilot desktop agent manager, and Web IQ search API for agents
Microsoft announced seven new AI models trained from scratch (not distilled from OpenAI), including the flagship MAI-Thinking-1 reasoning model and MAI-Transcribe-1.5, plus a 'Frontier Tuning' reinforcement learning approach for enterprise workflow training. GitHub released a desktop Copilot app designed to manage multiple parallel AI agents with isolated git worktrees and bidirectional canvases. Microsoft also launched Web IQ, an agent-native Bing-powered grounding API already powering search in Copilot and ChatGPT, running 2.5x faster than alternatives with lower token costs. The roundup also covers Nous Research's Hermes Desktop cross-platform agent app, Alibaba's Qwen3.7-Plus multimodal model, and OpenAI's role-specific Codex plugins.