10 posts tagged MCP. ← Back to all posts
Agents arrive at each session without persistent state. Externalization mechanisms differ in what they capture: intent, declaration, structure, execution. Treating them as one bucket obscures the architectural choice in each. This post separates the four artifacts produced by the work shipped in 2026.
A few hours after writing about workflow-atlas as a context layer for Claude, I shipped the execution-layer companion. mcp-tape records every MCP JSON-RPC call your agent makes; mcpreplay.dev lets you read what happened. Open source, MIT, alpha.
Anthropic launched Managed Agents (beta) — composable APIs for cloud-hosted vertical agents. The first one I'm building on top is Compliance-bot: hand it a robot repo, it returns five signed attestation packets ready for EU AI Act submission. Why this is the open-core wedge robot-md needed.
v1.4.0 ships three independent feature tracks that, together, change what registering a new piece of hardware feels like: hot-plug detection at the OS level, an announce/confirm UX over MCP that lets Claude talk the operator through the bind, and the first half of SP6's self-attestation loop — robots now sign their own spatial-intelligence scores with their RCAN apikey.
An introduction to the OpenCastor ecosystem — a single ROBOT.md manifest, runtimes in Python and TypeScript, an MCP server, a Claude Code plugin marketplace, a public registry with EU AI Act-aligned compliance endpoints, and the consumer-tier UX I've been focused on lately. Built almost entirely with Claude Code, on a SO-ARM101 arm I call bob.
A two-command install that wires your robot's ROBOT.md into any Claude Code session as structured MCP resources. Also: what RCAN and the Robot Registry Foundation do, and why this plugin belongs in Anthropic's hosted MCP catalog.
External systems need to drive robots too. The obvious move is to put robot-md-mcp behind an HTTPS frontdoor. It works. It's the wrong shape. Here's the right one — a BYOK Claude Agent SDK dispatcher that runs on the robot host.
What happens when you give Claude Code direct, LoA-gated, RCAN-signed access to a physical robot via the Model Context Protocol.
OpenCastor now exposes a full MCP server with LoA-gated tools. Any AI agent — Claude Code, Codex, Gemini, cron jobs — can command and observe robots through a single interface. Here's why we built it, and how compliance requirements shaped the design.
A systematic audit of the OpenCastor RCAN v2.2 implementation surfaced 17 gaps across CLI, gateway, SDKs, and the Flutter client. We closed all of them.