Releases·Tool Release
MLflow MCP Server: Debug, Analyze, and Annotate Traces from Any AI Assistant
MLflow's MCP Server exposes tools for AI assistants to read bulk experiment data — logs, traces, artifacts, prompts — and to write, update, and log feedback on traces, going beyond the read-only, paginated MLflow UI. This gives practitioners assistant-driven debugging and annotation of LLM traces directly from any MCP-compatible client.
5 engagement1 sourceTue, Sep 8, 2026, 06:02 PM
AI Overview
- MLflow MCP Server reads and writes MLflow tracking server data, since the UI is read-only with paginated trace views
- Works via Fastmcp Python client programmatically or wired to Claude, Cursor, and VSCode for natural-language reads and writes
- A cookbook and notebook with code examples for both access methods ship with the tutorials
Related
Social·Community
Serj Smorodinsky and Brett Kennedy to lead Oct 3 DSPy workshop on eval-driven LLM development with MLflow
A 3-hour live workshop on October 3, led by AI engineers Serj Smorodinsky and Brett Kennedy, teaches moving from hand-tuned prompts to structured, measurable LLM development using DSPy signatures/modules and MLflow tracking. The Reddit post promoting it targets practitioners running RAG in production who want reproducible evaluation instead of ad-hoc quality checks.
A 3-hour live workshop on October 3, led by AI engineers Serj Smorodinsky and Brett Kennedy, teaches moving from hand-tuned prompts to structured, measurable LLM development using DSPy signatures/modules and MLflow tracking. The Reddit post promoting it targets practitioners running RAG in production who want reproducible evaluation instead of ad-hoc quality checks.
2 engagement1 sourceredditWed, Sep 30, 2026, 04:02 AM
Products·launched
Natural language querying of agent traces via Databricks Genie
This is a setup where an agent's production traces are logged to MLflow and queried using Databricks Genie, which translates natural language questions into SQL. It solves the problem of manually writing pandas/SQL to analyze agent performance, making it easier for developers to get insights like slowest traces or p95 latency.
This is a setup where an agent's production traces are logged to MLflow and queried using Databricks Genie, which translates natural language questions into SQL. It solves the problem of manually writing pandas/SQL to analyze agent performance, making it easier for developers to get insights like slowest traces or p95 latency.
7 engagement1 sourceredditFri, Aug 14, 2026, 10:55 PM
Releases·Tool Release
Simon Willison documents steps to add custom MCP server to Claude and ChatGPT
Simon Willison published a TIL guide detailing the multi-step process for connecting a custom MCP server to the standard chat interfaces of Claude and ChatGPT. The guide covers the necessary configuration and setup steps for developers to integrate custom tools via the Model Context Protocol.
Simon Willison published a TIL guide detailing the multi-step process for connecting a custom MCP server to the standard chat interfaces of Claude and ChatGPT. The guide covers the necessary configuration and setup steps for developers to integrate custom tools via the Model Context Protocol.
0 engagement1 sourcerssWed, Jul 29, 2026, 12:13 AM
Social·Community
Community requests native MCP server for Claude Code to enable bidirectional tool control
A Reddit user proposes that Anthropic add an opt-in native MCP server to Claude Code, allowing other applications to control Claude Code as a tool. The feature would enable starting sessions, submitting tasks, reading progress, and requesting code reviews, with security defaults like localhost-only access and audit logging.
A Reddit user proposes that Anthropic add an opt-in native MCP server to Claude Code, allowing other applications to control Claude Code as a tool. The feature would enable starting sessions, submitting tasks, reading progress, and requesting code reviews, with security defaults like localhost-only access and audit logging.
6 engagement1 sourceredditMon, Jul 13, 2026, 12:19 PM
Products·building
Unified workspace for debugging multi-step AI workflows
A developer tool that consolidates traces, prompts, logs, and metrics into a single interface for investigating AI workflow executions. It organizes data hierarchically as Projects -> Sessions -> Runs -> Events, aiming to reduce context switching. Built for developers building AI products or agentic pipelines.
A developer tool that consolidates traces, prompts, logs, and metrics into a single interface for investigating AI workflow executions. It organizes data hierarchically as Projects -> Sessions -> Runs -> Events, aiming to reduce context switching. Built for developers building AI products or agentic pipelines.
21031 engagement4 sourcesredditSat, Jul 25, 2026, 08:11 AM