Open-Source extensible-mcp Adds Zero-Trust Policy Control to LLM Tool Calls
Fearless-Might-5439 · reddit · 2026-10-29
SenteLabs open-sourced extensible-mcp, an MCP proxy for securing LLM-powered applications: it loads new MCP servers dynamically without polluting the model's context and applies zero-trust policy controls to every tool call.
- Policies and credentials live outside the LLM's purview; policy languages supported include Rego, CEL and newly added Cedar
- Tool definitions from MCP servers are stored in a queryable vector store, so the LLM only sees tools it uses
- Key design: distinguishing rights from intents — an agent acting on your behalf should follow your intent, not do anything you could do just because of a hallucination
Already used in the authors' Open Executive project; Apache-2.0 licensed.
More from coding & agent
- Grok voice agent breaks down, sparking rant against renting other people's agent harnesses — alexcovo_eth · 2026-10-29
- A silent skip counted as success: rebuilding a bot lane that can't go quiet unnoticed — petrusenko_max · 2026-10-29
- Independent dev: 'I just talk to Claude and the feature gets implemented' — evielync · 2026-10-29
- Dev argues subagent fleets are pointless with Claude Opus 5.5 — one-on-one prompting wins — ryunuck · 2026-10-29
- The "Brain Sandwich": a three-step workflow for using AI without atrophying your engineering judgment — bibryam · 2026-10-29
- Nolan Lawson benchmarks the AI cost of JS frameworks: results mostly similar — hichaelmart · 2026-10-29