Video wird geladen...

Video konnte nicht geladen werden

Zur Startseite

Hey MCP developers! Let’s talk about something broken 🧠🛠️ Agents are calling tools left and right. But what happens inside those tools? No traces. No visibility. No security. Just a black box. 🕋 The observability gap is real. Let's fix it together with observable[.]tools

40,739 Aufrufe • vor 1 Jahr •via X (Twitter)

12 Kommentare

Profilbild von Weights & Biases
Weights & Biasesvor 1 Jahr

We believe observability shouldn’t be a bolt-on. It should be a first-class citizen in the agent stack. So we’re launching initiative to bring full-stack tracing to MCP tools using OpenTelemetry. Think: from black boxes → glass boxes. ⎚

Profilbild von Weights & Biases
Weights & Biasesvor 1 Jahr

To achieve this, we propose a combining OTel (open telemetry) into the official MCP protocol via a spec RFC. Combining two vendor neutral, open protocols to enable easy observability for MCP developers and tool makers (both client and server). Full proposal details on Github:

Profilbild von Weights & Biases
Weights & Biasesvor 1 Jahr

We want to build this ecosystem together. W&B is extending an invitation to our friends from the observability industry. So... 👀 @LangChainAI, @braintrustdata, @pydantic, @arizeai, @rungalileo, @AgentOpsAI & others. Will you join this effort with us, to enable observability for the million incoming MCP developers?

Profilbild von Weights & Biases
Weights & Biasesvor 1 Jahr

Most importantly—this is a call to YOU, the MCP tool developers. ✅ Read the manifesto 💬 Weigh in on the spec RFC 🛠️ Start building tools with observability baked in The agentic future demands transparency. Let’s build it right at

Profilbild von SecBriefs | Making Cybersecurity Simple
SecBriefs | Making Cybersecurity Simplevor 1 Jahr

Quantum computers can break today’s encryption in seconds.🔑 Quantum tech will reshape our digital lives. Governments & hackers are preparing for the quantum era. How about you?🛡️ Don’t get left behind!🧠 Cybersecurity Dictionary for Everyone can help:

Profilbild von William Horton at @PyTexas
William Horton at @PyTexasvor 1 Jahr

Why wouldn’t MCP servers just use whatever observability stack the org who built/hosts the server is using? MCP is just a certain kind of API

Profilbild von Weights & Biases
Weights & Biasesvor 1 Jahr

Totally fair q, William. We believe agent developers need end-to-end visibility across tools. If each server logs to its own stack, there's no unified trace tree. Our approach keeps server setup simple and gives the client full control of observability.

Profilbild von Fadi "Ali"
Fadi "Ali"vor 1 Jahr

Absolutely! Let's open those black boxes and bring transparency to AI tools. Count me in to tackle the observability gap!

Profilbild von Weights & Biases
Weights & Biasesvor 1 Jahr

🫡

Profilbild von web3MATO 🐦‍🔥
web3MATO 🐦‍🔥vor 1 Jahr

Chain abstraction is a game-changer, but adoption is the real test! 🚀 Curious—how does @wandb plan to tackle scalability & cross-chain UX? Looking forward to seeing this evolve!

Profilbild von Dylan Heagy
Dylan Heagyvor 1 Jahr

What do you think is the biggest barrier to adoption for this?

Profilbild von Weights & Biases
Weights & Biasesvor 1 Jahr

Hey Dylan, great question! We've outlined a few issues in the MCP RFC we posted, namely - While this isn't in the MCP spec, there isn't a standard way for clients and servers to show the developer a combined trace tree - For servers, authentication into Otel backends will be an issue, so we propose a way to have the Client receive and organize all the traces - As with every protocol, even if our proposal is accepted, adoption into the various SDKs (Python, TypeScript, C# etc) will take time But we are committed to make sure MCP will be fully observable!

Ähnliche Videos

HERMES AGENT WITHOUT TOOLS IS A CHATBOT. WITH THEM IT BUILDS 3D TOWERS IN BLENDER, CHECKS STOCK PRICES, AND DRIVES VS CODE. tonbi JUST DROPPED THE FULL GUIDE. module 6 of his 10-part Hermes masterclass. best breakdown of the tool layer anyone has published. what you need to know: TOOLS vs SKILLS vs MCP skills = instructions (markdown, loaded into context) tools = callable functions (Python, agent emits call, Hermes executes) MCP = adapters to external systems (Blender, Stripe, Linear, Notion) every tool has three parts: → the function (does the real work) → the schema (what the model sees to decide when to call it) → the registry (makes the tool exist in the agent) the model never runs the function directly. it emits a structured request (tool name + JSON args). Hermes executes and returns the result. if the tool fails, the error goes back as JSON. the agent recovers instead of crashing. TOOL SETS CONTROL THE SURFACE hermes chat --tool-sets web # only web tools loaded. no files, no terminal. hermes chat --tool-sets safe # read-only: web search, vision, image gen. # no file writes. no terminal. no code execution. mid-session: /tools enable video /tools disable terminal or toggle in the dashboard: hermes dashboard → Skills → Tool Sets MCP SERVERS two transport types: STDIO (local subprocess) or HTTP (remote endpoint). add a server: hermes mcp # or ask: "add the MCP server for Blender" filter tools with include/exclude per server. keep only what you trust. security: → OAuth 2.1 PKCE (no long-lived tokens in config) → package scanning via api.osv. dev before launch → all MCP calls go through approval gates HERMES AS MCP SERVER hermes mcp serve exposes 10 tools via FastMCP. connect VS Code Copilot or Cursor to your running Hermes instance. BUILD YOUR OWN TOOL He built a stock price tool live: → Python function calling Finnhub API → schema with name, description, parameters → registered in tool_sets.py → agent calls it automatically when relevant any repeating API call in your workflow can become a native tool. full Hermes architecture deep-dive in the article 👇

YanXbt

32,680 Aufrufe • vor 1 Monat