json
30 articles about json in AI news
MCP Becomes USB for AI: 3 Primitives, JSON-RPC 2.0, 50+ Servers
Anthropic's MCP standardizes AI tool connections via JSON-RPC 2.0 with three primitives. Over 50 community servers exist, making it the USB for AI.
How Structured JSON Inputs Eliminated Hallucinations in a Fine-Tuned 7B Code Model
A developer fine-tuned a 7B code model on consumer hardware to generate Laravel PHP files. Hallucinations persisted until prompts were replaced with structured JSON specs, which eliminated ambiguous gap-filling errors and reduced debugging time dramatically.
How to Build Interactive Apps with Claude Code Using JSON as a Bridge
A developer built a music maker where Claude generates beats via JSON, showing how to create collaborative AI tools without complex APIs.
Stop Clicking 'Approve': A .claude/settings.json Template for 80% Fewer
A practical guide to configuring Claude Code's permissions file to auto-approve routine development commands, speeding up your workflow without sacrificing safety.
How to Cut Agent Token Waste: CLI Over GraphQL + Server-Pushed Hints
Replace raw GraphQL with typed CLI commands to eliminate JSON assembly errors, then add server-pushed hints via MCP to prevent judgment failures. Your agent burns 1,500+ tokens per operation otherwise.
Build a Zero-Dependency MCP Server
Build a zero-dependency MCP server in 50 lines of Python to give Claude Code direct file access. Register it in claude.json and skip SDK bloat.
SSL: Structured Skill Language Boosts Skill Discovery MRR to 0.707
Researchers propose SSL, a three-layer typed JSON representation for AI agent skills, replacing unstructured SKILL.md prose. Using an LLM normalizer, SSL improves Skill Discovery MRR from 0.573 to 0.707 and Risk Assessment macro F1 from 0.744 to 0.787 on a newly released 6,184-skill corpus.
DharmaOCR: New Small Language Models Set State-of-the-Art for Structured
A new arXiv preprint presents DharmaOCR, a pair of small language models (7B & 3B params) fine-tuned for structured OCR. They introduce a new benchmark and use Direct Preference Optimization to drastically reduce 'text degeneration'—a key cause of performance failures—while outputting structured JSON. The models claim superior accuracy and lower cost than proprietary APIs.
Claude Code Hooks: How to Auto-Format, Lint, and Test on Every Save
Configure hooks in .claude/settings.json to run prettier, eslint, and tests automatically, ensuring clean code without manual intervention.
BloClaw: New AI4S 'Operating System' Cuts Agent Tool-Calling Errors to 0.2% with XML-Regex Protocol
Researchers introduced BloClaw, a unified operating system for AI-driven scientific discovery that replaces fragile JSON tool-calling with a dual-track XML-Regex protocol, cutting error rates from 17.6% to 0.2%. The system autonomously captures dynamic visualizations and provides a morphing UI, benchmarked across cheminformatics, protein folding, and molecular docking.
What Cursor's 8GB Storage Bloat Teaches Us About Claude Code's Clean Architecture
A deep dive into Cursor's scattered 8GB local storage reveals why Claude Code's ~/.claude/projects/*.jsonl approach is better for developers.
How This Developer Built a Production-Ready RAG System with Claude Code in One Weekend
A developer used Claude Code to create a structured JSON-to-PDF knowledge base with 105 quotes, demonstrating how to build RAG-ready datasets faster than ever.
3 MCP Gateway Security Gaps LiteLLM's Audit Found (And How to Fix Them in
LiteLLM's audit revealed 3 MCP gateway gaps: fail-open resolver, unpinned servers, opt-in least-privilege. Fix them in Claude Code with version pinning and allowed_tools.
Google Cloud Joins MCP: How to Connect Claude Code to BigQuery
Google Cloud's MCP server lets Claude Code query BigQuery and manage GCS directly. Install it with `claude mcp add google-cloud` and authenticate.
Pipedrive Ships Native MCP Server, CRM Joins AI Agent Protocol
Pipedrive launched a native MCP server for CRM agent access. The move follows Microsoft and X in adopting Anthropic's open standard, which crossed 13,000 servers in June.
Stop Hardcoding Model Lists: Use Discovery-Driven MCP to Cut Token Bloat 40%
Switch from hardcoded MCP tool schemas to discovery-driven tools like nvidia_list_foundation_models. Your agent queries available models dynamically, cutting token bloat and adapting to infrastructure changes in real-time.
Argus for Claude Code: Cut Repetitive Workflow Token Usage by 80% with
Argus enforces deterministic workflows for Claude Code, cutting token usage ~80% by replacing AI reasoning with structured execution paths. Install in minutes.
Stop Testing Skills Once: Use Caliper's pass@k to Measure What Actually
Caliper is a lightweight harness that runs Claude Code skills k times, scores them with pass@k, and compares against a no-skill baseline so you know if your skill actually helps.
Two Bash Hooks Cut an Anthropic Bill From $312 to $156
A developer halved their Anthropic bill from $312 to $156 by adding two Bash hooks to Claude Code that intercept retry spirals and block dangerous deploys.
Caliper: Run Your Claude Code Skills k Times and Get a pass@k Score That
Caliper gives Claude Code users a pass@k reliability score for skills, with a baseline delta showing if the skill beats the base agent. Install via pipx or npx.
Expose pgvector as an MCP Server: From Hardcoded RAG to Reusable Tool Server
Wrap pgvector search in FastMCP to create a reusable MCP server. Any LLM client—including Claude Code—can then query your vector database without hardcoded integrations.
My Day-One Claude Code Setup: A Workflow That Saves 15 Minutes Per New Project
A repeatable workflow with CLAUDE.md, project templates, and MCP servers saves 15 minutes per project and reduces setup errors by 40% for Claude Code users.
MCP Server Versioning: How to Avoid Breaking All Your AI Clients (Like I
Stop breaking AI clients with MCP schema changes. Use query param versioning (?v=2) — it works with every MCP client, requires no code changes, and lets old and new versions coexist seamlessly.
Propel Ships First Production MCP Server for PLM
Propel Software launched the first production MCP server for PLM, connecting LLMs to live product data. No competitor has matched this open-protocol approach.
Building a Production-Ready Snowflake MCP Server: A Practical Guide
A technical guide details building a production-ready Snowflake MCP server with OAuth 2.0, schema filtering, and rate limiting for enterprise AI agents.
Why Your CLAUDE.md Needs a 'No npm install' Rule for Open Source Repos
Add a `# In open source repos, never run npm install or pip install without asking first` rule to CLAUDE.md. This prevents Claude Code from executing untrusted code, saving tokens and protecting your system.
World Model MCP: Memory Layer That Cut SWE-bench Repeat Mistakes by +10.2 Points
World Model MCP adds a temporal knowledge graph to Claude Code that learns from corrections, prevents repeated mistakes, and re-injects context after compaction — proven with +10.2 pts on SWE-bench.
How to Build a Native MCP Server
Native MCP servers embed the Model Context Protocol directly into your CMS, letting Claude Code act like a trusted team member with full permission enforcement. No more copy-paste.
Stop Writing Rules in CLAUDE.md—Use PreToolUse Hooks for Guaranteed Enforcement
Replace CLAUDE.md rules with PreToolUse hooks in Claude Code for guaranteed enforcement. Hooks run as shell code, blocking dangerous commands like deploys or migrations outside the model's control.
x402 Lets Agents Pay Per API Call — No Wallet Required
x402 lets AI agents pay per API call via USDC micropayments. The Stall now adds a fiat rail: buy credits with a card from $5/100 calls.