Daily brief · Saturday, August 1, 2026

Cursor Strips Per-Request Dollar Cost Tracking for Non-Enterprise Plans

Cursor's unexpected removal of dollar-based usage metrics forces engineering teams and individual developers to adjust their cost-tracking workflows.

12 min read6 storiesAI-assisted · editor-reviewed

0 / 6read

Category banner for the lead story · Agents & MCP

The brief in 30 seconds

Agents & MCPMCP 2.0 simplifies tool calling down to a single HTTP POST request.
Models & researchDeepSeek V4-Flash-0731 provides 304B parameters at $0.14/M input and $0.27/M output pricing.
Local LLMsTreat local LLMs as targeted assistants for technical summarization and note synthesis.

Agents & MCP
2 min · Simon Willison
Inspect and Query Stateless Model Context Protocol Endpoints with mcp-explorer and llm-mcp-client
Following the Model Context Protocol 2.0 stateless HTTP specification, new developer CLI tools simplify probing and interacting with Model Context Protocol servers without maintaining server-side session state. Developers can now inspect tool schemas instantly via uvx and query tools directly inside terminal LLM workflows using alpha llm-mcp-client.
Why it matters. Stateless MCP eliminates the need for managing server session state or complex connection routing, making it significantly easier to build and audit agent tool integrations.
Full context
Models & research
2 min · Simon Willison
Maximize DeepSeek V4-Flash Output Quality Using High Reasoning Effort Flags
DeepSeek V4-Flash-0731 offers 304 billion parameters at $0.14/M input and $0.27/M output pricing. Setting the reasoning effort parameter to high substantially improves output quality on complex prompts compared to default reasoning levels.
Why it matters. DeepSeek V4-Flash punches above its weight according to Artificial Analysis, ranking ahead of MiniMax M3 (428B), making output tuning via reasoning flags critical for cost-effective performance.
Full context
Local LLMs
2 min · Mastodon
Practical Workflows for Running Local Models with Ollama
Running local language models with Ollama provides a private environment for prompt testing, technical documentation summarization, and content structuring. Instead of replacing research, local models act as offline assistants without incurring API fees.
Why it matters. Developers can integrate local inference into daily workflows to summarize long changelogs and test prompts without API latency or token costs.
Full context
Tools & releases
2 min · Hacker News
Aurora: Open-Source Self-Hosted AI Gateway in Go for Multi-Provider Routing
Aurora is an Apache 2.0 Go-based AI gateway offering OpenAI and Anthropic compatibility across 14 providers. It features provider pooling, prompt/response caching, guardrails, and token savings in a single binary.
Why it matters. Swap LLM API base URLs to instantly gain failover, unified auditing, semantic caching, and token compression without updating application code.
Full context
Agents & MCP
2 min · Hacker News
Wienerdog: Shared File-Based Memory and Self-Improving Skills for Claude Code and Codex
Wienerdog is a serverless, zero-telemetry CLI tool that creates a shared Markdown memory vault and auto-improving skills across Claude Code and Codex CLI. It features a nightly dreaming process that condenses session histories into Obsidian-compatible PARA notes.
Why it matters. Switching between AI coding tools usually resets context. Wienerdog maintains persistent, file-based memory shared across tools locally on your machine without running extra daemons or sending telemetry.
Full context
Tools & releases
2 min · Hacker News
Cursor Removes Dollar Cost Data from Usage Dashboard and CSV Exports
Cursor updated its dashboard to replace dollar amounts with token counts for all self-serve individual and Teams plans. Cost fields in CSV exports were set to 0.0, and API cost fields were zeroed out or omitted retroactively, breaking custom cost-tracking scripts.
Why it matters. This change disrupts self-serve teams and individual developers who rely on real-time API cost tracking to manage budgets and monitor model usage.
Full context

One thing to try

Cursor Strips Per-Request Dollar Cost Tracking for Non-Enterprise Plans

The brief in 30 seconds

Inspect and Query Stateless Model Context Protocol Endpoints with mcp-explorer and llm-mcp-client

Maximize DeepSeek V4-Flash Output Quality Using High Reasoning Effort Flags

Practical Workflows for Running Local Models with Ollama

Aurora: Open-Source Self-Hosted AI Gateway in Go for Multi-Provider Routing

Wienerdog: Shared File-Based Memory and Self-Improving Skills for Claude Code and Codex

Cursor Removes Dollar Cost Data from Usage Dashboard and CSV Exports

Install Ollama locally and pull an open-weights model like Qwen or Llama.

Cursor Strips Per-Request Dollar Cost Tracking for Non-Enterprise Plans

The brief in 30 seconds

Inspect and Query Stateless Model Context Protocol Endpoints with mcp-explorer and llm-mcp-client

Maximize DeepSeek V4-Flash Output Quality Using High Reasoning Effort Flags

Practical Workflows for Running Local Models with Ollama

Aurora: Open-Source Self-Hosted AI Gateway in Go for Multi-Provider Routing

Wienerdog: Shared File-Based Memory and Self-Improving Skills for Claude Code and Codex

Cursor Removes Dollar Cost Data from Usage Dashboard and CSV Exports

Install Ollama locally and pull an open-weights model like Qwen or Llama.