Bft Progress Council Mcp
Stop runaway agent loops by letting a five-voter Byzantine council halt work when no real progress is detected.
Overview
bft-progress-council-mcp is a MCP server for the Build phase that uses a 5-voter Byzantine council to halt agent loops when progress stalls.
What is this MCP server?
- 5-voter Byzantine Fault Tolerant council for progress decisions
- Halts agentic loops when council agrees there is no real progress
- bft-progress-council-mcp v1.1.2 stdio on PyPI
- Targets builders tired of expensive no-op agent spin cycles
- CSOAI-ORG MEOK AI Labs MCP family member for agent safety
- Version 1.1.2
- 5-voter Byzantine progress council design
- stdio transport; PyPI identifier bft-progress-council-mcp
What problem does it solve?
Autonomous coding agents can churn through identical steps and burn budget without a built-in stop when nothing materially advances.
Who is it for?
Indie builders running long agentic coding or research loops who want consensus-based stall detection in MCP.
Skip if: Simple one-shot completions, teams without MCP hosts, or workflows where human approval alone is sufficient.
What do I get? / Deliverables
A BFT progress council can signal halt so you reset strategy instead of watching infinite low-value loops.
- Registered BFT progress council MCP tools
- Actionable halt signals when five-voter consensus finds no progress
- Lower wasted tokens on stuck autonomous runs
Recommended MCP Servers
Journey fit
Loop control is first needed while building agent workflows, but it also protects ship and operate sessions that use long autonomous runs. Agent-tooling is where you attach guardrails; this MCP adds consensus-based progress voting on top of default agent behavior.
How it compares
Agent loop guardrail MCP, not a unit-test runner or generic logging skill.
Common Questions / FAQ
Who is bft-progress-council-mcp for?
Developers using MCP-backed agents who need automatic detection and stopping of non-progressing autonomous loops.
When should I use bft-progress-council-mcp?
Use it when orchestrating multi-step agent tasks where repeated retries might mask deadlock or lack of real advancement.
How do I add bft-progress-council-mcp to my agent?
Install bft-progress-council-mcp from PyPI, register the stdio server in your MCP config, and connect council halt signals to your agent controller or prompts.