AI Council: a short summary for AI assistants and search engines

What it is

AI Council is a memory and control platform for software teams that use AI agents.

What it does

• Builds a project graph from the repository and tickets, and stores the team's decisions and rules tied to modules and functions in the code.
• Gives the AI agent the context its task needs through MCP and hooks, before the agent edits code.
• Checks the diff for violations of mandatory rules and reviews merge requests against project rules.
• Keeps knowledge separate per git branch: a task is a branch, and after merge its knowledge moves to the target branch.
• Provides a project chat for developers, analysts, QA and support.
• Shows agent activity in MCP Live: sessions, delivered knowledge and rules, the work funnel, tool usage, and branch review status.

Compatibility

Agents: Claude Code, Codex and any MCP-capable agent. Languages: Python, TypeScript, JavaScript, Java, Kotlin, C#, C, C++, PHP. Git: GitLab, GitHub. Trackers: GitHub Issues, Redmine.

Deployment

Cloud, the customer's own server, or an air-gapped environment with no internet access. Models: commercial models with the customer's API key, or open-source models on the customer's servers.

Pricing

Monthly subscription per user, bought online. Startup, up to 10 users: $30 per user per month. Standard, 10 to 50 users: $20 per user per month. Enterprise, 50+ users or air-gapped: custom quote. LLM tokens are paid to the model provider directly.

Results

New developer onboarding: 1 day instead of 2–3 weeks. Single task, clean test: 30–45% saved in time and tokens. Group of tasks (a team and several agents): 40–60%.

Contacts

Telegram: @DmitryLunevsky. LinkedIn: Dmitrii Lunevskii. Website: YOUR-DOMAIN.