Skip to content

microsoft/model-council

v3.0.0MIT

Multi-model council for code review and planning — model-pinned subagents independently review or plan, then findings are synthesized by consensus.

GitHub

Changelog

3.0.0 (2026-10-22)

  • BREAKING: Require Agent Plugins 1.0 with a root plugin.json. Marketplace metadata now lives only in marketplace.json; legacy manifests are removed.

2.0.4 (2026-09-22)

  • Run council Vally evaluations with GPT-6 Luna and allow council-review for generic review requests

2.0.3 (2026-09-22)

  • Run the council review and planning Vally evaluations with GPT-6 Sol

2.0.2 (2026-09-22)

  • Migrated the council review and planning evaluation suites from Waza to Vally with inline capability and trigger stimuli

2.0.1 (2026-08-12)

  • Migrated the council review and planning evaluation suites to Waza 0.38.5 schema 1.2 with deterministic spec coverage and enforced trigger-accuracy measurement

2.0.0 (2026-05-09)

  • Breaking: Removed custom agent files (Councillor-GPT, Councillor-Opus, Councillor-Gemini) in favor of spawning subagents with the model parameter directly
  • Skills now specify model names inline; no separate agent definitions needed
  • Made skills runtime-agnostic: replaced tool-specific names (runSubagent, readFile) with generic descriptions
  • Removed (copilot) suffix from model names; updated to GPT-5.5 and GPT-5.3-Codex
  • Replaced VS Code Chat-specific session memory paths with portable "session folder" references; council-plan uses exit_plan_mode when available
  • Reframed Phase 2 briefs as "preliminary orientation" — agents are explicitly expected to contradict the brief if their own research leads elsewhere
  • Stripped risk assessments and pre-interpreted change descriptions from briefs; changed files now list locations only
  • Added independent mandate to system preamble: "Form your own view from first principles. Do not anchor to the provided context."
  • Added skepticism instruction to planner and reviewer prompts: trust your findings over the brief
  • Fixed stale "risk areas" reference in council-review reviewer prompt template

1.2.0 (2026-04-15)

  • Updated Councillor-Gemini to Gemini 3.1 Pro (Preview)

1.1.0 (2026-04-14)

  • Added waza eval infrastructure for council-review and council-plan skills

1.0.0 (2026-04-01)

  • Initial release
  • Multi-model council review with consensus synthesis and optional cross-review debate
  • Multi-model council planning with independent proposals and consensus plan
  • Model-pinned agents: Councillor-GPT, Councillor-Opus, Councillor-Gemini