One model gives one guess. A cross-vendor panel of up to 11 frontier models fans out in parallel and gives you the union of what they all see — the risks, edge cases, and approaches a single model misses.
AI Consultants convenes a panel of up to 11 frontier models — each with a distinct persona and a different vendor's blind spots — and fans your question out to all of them in parallel. No back-and-forth rounds, no voting on one answer. Every consultant answers once, independently.
Different model families miss different things. One model's blind spot is another's obvious flag. Fanning the question out and taking the union of what the panel collectively sees catches what any single model would have missed on its own.
The result: a coverage synthesis — the deduplicated union of every distinct point raised across the panel — so you see the full spread of risks, edge cases, and approaches, not just one model's take.
Built for developers who want more than a single opinion.
Gemini, Codex, Mistral, Cursor, Kimi K3, Claude, Qwen3, GLM, Grok 4.5, DeepSeek, and MiniMax. Each with a distinct role.
Every response is deduplicated and merged into the union of distinct points across the panel — the risks, edge cases, and approaches, not a single weighted recommendation.
Questions are classified by category and routed to the most relevant consultants. Security questions go to security experts first.
Premium, standard, and economy model tiers. Choose max_quality for critical decisions or fast for quick checks.
Budget enforcement, semantic caching, cost-aware routing, and response limits. Every figure states what it rests on — the provider’s own token counts where they exist, a local estimate where they don’t, and an explicit exclusion for credit-billed models.
Share large files without burning the consultants' context window. AST-based skeleton extraction, category-aware project tree, and PRIMARY/CONTEXT relevance tags let you point the panel at what matters.
Your question is categorized — architecture, security, performance, code review — and scored for complexity and intent. From that, routing picks which consultants in the panel see this specific question.
The relevant consultants are queried in parallel, independently. Each returns a structured response — summary, analysis, pros/cons, and a 1–10 confidence score — from its own persona and focus area.
Every response is deduplicated and merged into the union of distinct points across the panel — the risks, edge cases, and approaches any one model alone would have missed. Proven strongest on breadth questions; for a single-answer factual question, one strong model is usually enough.
Each consultant has a distinct persona that shapes their analysis. The invoking agent is automatically excluded to prevent self-consultation. The focused roster contains 11 supported consultants.
Each recipe is a complete control surface for one job. Paste it into your shell for a one-off run, or move the variables into ~/.config/ai-consultants/.env to make it your default.
Every consultant answers once in parallel, and the coverage synthesis lays out where they agree and where they don't — then names a winner.
ai-consultants --strategy majority \ "Choose REST or GraphQL for this API \ and justify one winner" # Trade-offs only: --strategy compare_only
Ping every selected consultant first and stop if fewer than three answer.
ENABLE_HEALTH_GATE=true \ HEALTH_GATE_TIMEOUT=30 \ QUORUM_MIN=3 \ QUORUM_ACTION=stop \ ai-consultants \ "Make a release recommendation"
Route economically, cap the session at $0.50, and keep partial results on stop.
ENABLE_BUDGET_LIMIT=true \ MAX_SESSION_COST=0.50 \ BUDGET_ACTION=stop \ ENABLE_COST_AWARE_ROUTING=true \ ai-consultants --preset cost-capped \ --strategy cost_capped \ "Find the smallest safe fix"
Override any older model stored in the user's Kimi configuration for this run.
ENABLE_KIMI=true \ KIMI_MODEL=kimi-code/k3 \ ai-consultants --preset balanced \ "Review this API design from a \ holistic perspective"
# Run directly — no install needed npx ai-consultants "How should I structure my auth system?" # With a preset npx ai-consultants --preset balanced "Redis or Memcached?" # Let the doctor recommend a preset for you (v2.13) npx ai-consultants doctor --suggest-preset --question "How can I prevent SQL injection?" # Scaffold persistent config at ~/.config/ai-consultants/ (v2.12) npx ai-consultants init # Detect your CLIs and keys, then write a full config (v2.22) npx ai-consultants configure # Run diagnostics npx ai-consultants doctor --fix # Install slash commands for Claude Code npx ai-consultants install # Keep the consultant CLIs up to date (v2.21) npx ai-consultants update-clis
# Install the skill curl -fsSL https://raw.githubusercontent.com/matteoscurati/ai-consultants/main/scripts/install.sh | bash # Ask your first question /ai-consultants:consult "How should I structure my auth system?"
# Clone and set up git clone https://github.com/matteoscurati/ai-consultants.git cd ai-consultants ./scripts/doctor.sh --fix # Run a consultation ./scripts/consult_all.sh "How should I structure my auth system?"
The automatic configurator detects the complete panel and preserves existing credentials. Presets choose the panel, strategies combine its answers, and every persistent parameter remains directly addressable.
# Detect installed CLIs and API credentials, then save an XDG config npx ai-consultants configure # Repeatable configuration for a coverage-focused architecture review npx ai-consultants configure \ --set DEFAULT_PRESET=balanced \ --set DEFAULT_STRATEGY=coverage # Guided setup, exhaustive review, and parameter discovery npx ai-consultants configure --interactive npx ai-consultants configure --advanced npx ai-consultants configure --show-parameters npx ai-consultants configure --dry-run
| Preset | Consultants | Use Case |
|---|---|---|
| minimal | 2 (Gemini + Codex) | Fast, cheap |
| fast | 2 + economy models | Quick checks |
| balanced | 4 (Gemini, Codex, Mistral, Cursor) | Standard consultations [default] |
| medium | 4 + standard models | General questions |
| thorough | 4 | Comprehensive analysis |
| cost-capped | 3 + budget cap | Minimal API costs |
| security | 4 + security-first routing | Security reviews |
| high-stakes | 5 of 11 | Maximum rigor for critical decisions |
| max_quality | 8 of 11 + premium models | Critical decisions |
| Strategy | Description |
|---|---|
| coverage | Deduplicated union of every distinct point across the panel (default) |
| majority | Most common answer wins |
| risk_averse | Weight conservative responses higher |
| security_first | Prioritize security considerations |
| cost_capped | Prefer simpler, cheaper solutions |
| compare_only | No recommendation, just comparison |
# Panel and synthesis DEFAULT_PRESET=balanced # minimal | balanced | thorough | high-stakes | security | fast DEFAULT_STRATEGY=coverage # coverage | majority | risk_averse | security_first | cost_capped | compare_only ENABLE_SYNTHESIS=true # Trust the panel size ENABLE_HEALTH_GATE=false # Adds one small live ping per selected consultant QUORUM_MIN=2 QUORUM_ACTION=warn # warn | stop # Cost ceiling ENABLE_BUDGET_LIMIT=false MAX_SESSION_COST=1.00 BUDGET_ACTION=warn # warn | stop