AI Expert Council at Your Fingertips
Every decision gets analyzed by multiple AI experts who disagree, debate, and synthesize — so you see every angle before you commit. Transparent reasoning. Customizable models. Your call.
BUILD WITH META COUNCIL
Create a scoped mc_ key in Settings, then use REST from any shell or connect an MCP-compatible agent. Authenticated API and MCP runs stay private.
Start a private council run.
curl -sS -X POST https://meta-council.com/api/v1/councils/run -H "Authorization: Bearer mc_your_key_here" -H "Content-Type: application/json" -d '{"query":"Should we expand into Europe?","panel":"default"}'Give your agent the council as a tool.
claude mcp add --transport http meta-council https://meta-council.com/mcp --header "Authorization: Bearer mc_your_key_here"Each expert's full reasoning is visible — their perspective, their concerns, their dissent. No black boxes. You see exactly how the recommendation was formed.
290+ specialized agents across finance, legal, engineering, healthcare, and more. They analyze independently, then a synthesis delivers a clear conclusion with risk matrix and action plan.
Choose which experts to consult. Set how much weight each opinion carries. Pick your model. Run on our servers or your own. The council works for you, not the other way around.
Describe the decision you're facing — the more context, the sharper the analysis.
Multiple AI experts analyze your query independently. They disagree. That's the point.
Not a framework — a clear recommendation with the reasoning, risks, dissent, and action plan to back it up.
Single-model AI gives you one perspective. Complex decisions need many.
One perspective. No dissent. No way to know what it missed. Hallucinations go unchecked. You get an answer but no way to verify the reasoning.
Multiple experts cross-check each other. Dissenting views surface blind spots. Research shows 30-40% fewer errors through multi-agent validation. Full reasoning visible.
The research backs this up. Recent studies show that multi-agent AI systems significantly reduce hallucination rates compared to single-model approaches. When multiple agents independently analyze a problem and cross-validate each other's outputs, factual errors are caught before they reach the user. One 2026 study found that structured multi-agent deliberation reduced hallucination rates by up to 30-40% across reasoning tasks. Read the research →
Expert AI Agents
Decision Panels
Workflow Pipelines
Reasoning Visible
Your queries contain your most sensitive strategic thinking. Meta Council keeps them under your control.
The entire platform — app, models, data — runs on your own servers. Free-tier queries use our self-hosted GPU inference with no third-party API calls. Enterprise customers can deploy on-premise.
All connections secured with TLS via Cloudflare. API credentials are stored server-side and never exposed to the browser after saving.
Every query, every agent response, and every synthesis is logged with timestamps. You can see exactly what the AI considered and why it recommended what it did.
Queries processed by our self-hosted models are never sent to third-party APIs and never used to train models. Delete your history anytime.
Go beyond one-shot queries. Run multi-step workflows — TDD coding, research methodology, due diligence, incident postmortems — where each step builds on the last and you review in between. Try Workflows →
Export any synthesis as a branded PDF for board presentations. Keep the run private, or explicitly publish a non-confidential session before sharing a link others can open.
Real output from a 3-agent Default Panel analyzing an M&A decision
Our B2B SaaS company with $12M ARR, 85% gross margins, and 140% NDR has received two acquisition offers. After careful analysis, we recommend accepting the strategic acquirer's offer of $180M cash plus an 18-month earn-out tied to 30% revenue growth. This decision aligns with our current profitability and provides a clear path forward without the uncertainties associated with the PE firm's longer-term equity stake. The founding team's split will be addressed through robust communication and negotiation strategies.
The $180M cash offer provides immediate financial stability and a strategic fit for our business model. The 18-month earn-out tied to 30% revenue growth acts as a buffer against potential revenue shortfalls while maintaining some level of control and independence compared to the PE firm's equity stake.
The PE firm's $150M offer with 40% management rollover equity targeting 3x in 5 years introduces significant long-term uncertainty. Given the founding team split and our current profitability, the strategic acquirer path offers a clearer and lower-risk trajectory.
| Risk | Likelihood | Impact | Raised By | Mitigation |
|---|---|---|---|---|
| Revenue Growth Pressure Earn-out tied to 30% growth may be challenging |
Medium | Medium | Strategist, Risk Analyst | Ensure robust revenue growth strategies and clear communication with the acquirer. |
| Founding Team Split Internal conflicts if not resolved promptly |
Medium | High | Devil's Advocate | Develop a comprehensive transition plan and engage in detailed negotiations to address the split. |
| Market Perception Accepting lower cash value vs. PE upfront |
Low | Medium | Devil's Advocate | Control the narrative; position the deal as strategic alignment rather than a discount. |
Assess the feasibility of achieving 30% revenue growth within 18 months. Review market conditions, product development pipeline, and sales strategies. (Strategist)
Engage in detailed negotiations with both parties to secure more favorable terms. Leverage the competing offers. (Devil's Advocate)
Create a comprehensive plan to address the founding team's split and ensure smooth operations during and after the transition. (Devil's Advocate)
Present the decision and plans to the board for final approval. Include the full risk matrix and agent reports. (Risk Analyst)
100% Free
No credit card required. Free forever.
$49/month
Try free for 7 days. Cancel anytime.
No charge for 7 days. $49/mo after. Cancel anytime before.
Custom Pricing
Run Meta Council entirely on your own machine. Your queries never leave your computer. No cloud, no API calls, no data sharing.
One command. Runs the web app + local AI model in containers. Works on any machine with Docker.
docker compose up
Native app for Mac, Windows, Linux. Double-click to launch. Auto-detects local Ollama for free inference.
Coming soon — contact us for early accessAir-gapped deployment. No internet required. Run Qwen 2.5 or Llama locally with your own GPU.
Requires: 8GB+ RAM, NVIDIA GPU optionalReady to make better decisions?
Enterprise & InvestorsScope a question, preview the agent research workflow, and request provider-returned source/citation records for independent verification by qualified counsel or the responsible authority.
Check the law requests legacy provider source/citation records and attaches a proposed downstream agent, source, evidence, and synthesis plan. The jurisdiction value is only a provider request hint; neither it nor returned records prove corpus coverage, exact citation identity, relevance, or authority. The plan is not executed. See what Themis says requests a versioned, neutral, evidence-linked research packet and is disabled until the unavailable verified contract exists. Legacy verdicts and answer prose are withheld. Neither action runs a legal consultation, gives advice, recommends an action, or makes a decision.
Meta Council and Themis provide informational research output only. They do not provide legal advice, recommendations, consulting services, legal opinions, or decisions, and they do not replace the independent judgment of a licensed attorney or responsible authority. No attorney-client relationship or privilege is created. Do not treat a model response, research plan, citation, quotation, coverage indicator, or freshness date as a determination that conduct is lawful, compliant, required, prohibited, or safe.
Laws, regulations, cases, administrative guidance, adopted model codes, local amendments, collective-bargaining terms, league and governing-body rules, and enforcement interpretations change and may conflict. Corpora can be incomplete, stale, incorrectly classified, or missing controlling local and fact-specific authority. Model quality, tools, jurisdiction detection, source availability, and supplied facts materially affect the output. Verify every source and its current effective status, and consult qualified counsel plus the applicable regulator, court, code official or authority having jurisdiction, governing body, or other competent authority before deciding or acting.
Submit a query to your expert panel for multi-perspective analysis and synthesis
1. Choose a panel of expert AI agents below
2. Ask your question — the more context, the better the analysis
3. Review the synthesis — see every expert's opinion, perspective, risks, and recommended actions
Applies to every expert and the synthesis. Choose Sonnet 5 for your next analysis, or Haiku 4.5 for lower-cost work. Both quick picks use your Anthropic API key from Settings.
Or try a sample query:
| Risk | Likelihood | Impact | Raised By | Mitigation |
|---|
Individual AI advisors, each with a distinct persona and analytical framework. Group them into Panels to convene a council.
Encode your own domain expertise into an AI advisor — no code required.
Encode your domain expertise into an AI agent. No code required — just describe how you think, what frameworks you use, and what questions you ask.
Write the analytical process, decision criteria, and expert persona.
Add your agent to custom panels alongside other experts.
List your agent for others to acquire. You keep 70% of every sale.
A panel is a curated group of Agents that deliberate together. Configure panels for different decision domains.
Describe your task and we'll set up the right workflow with the right experts.
Or choose a template:
Loading schedules...
Browse, customize, or build from scratch
Templates published by other builders on the platform
No nodes yet. Click "Add Node" to start building.
No nodes added yet.
Define how expert opinions are weighted, combined, and resolved into decisions
This strategy is injected into the synthesis step. It controls how conflicting opinions are resolved, how weights are applied, and what the output format looks like.
Create a FREE account to build and manage your own synthesis strategies.
Discover, share, and sell custom agents and councils built by the community
No marketplace agents yet. Create and list your own agents to get started!
No marketplace councils yet. Publish a panel to share it with the community!
Create a FREE account to see your agents and councils.
One balance that pays every data provider, API and MCP server on the platform. 100 credits = $1.00.
Loading wallet…
Pay any listed provider from your credit balance — no new payment integration per vendor. Paying settles the bill; it does not grant data access or change how that source's data may be used.
Loading providers…
Every movement, newest first. Ledger rows are append-only — nothing here is ever edited or deleted.
Loading statement…
List an API, MCP server, dataset, service or yourself as a payee. New listings start as drafts and are invisible to others until you activate them.
Loading listings…
Only earned credits can be cashed out — credits you bought are spend-only. Filing a request fences the credits immediately; an operator releases the payout by hand.
Loading requests…
Turn an explicit option slate and ranked ballots into a deterministic, reviewable preference. A result never records a Human Decision or starts work.
Past council sessions and workflow runs
Plan, assign, and track work for humans and agents — the same board agents use via tools, API, and MCP.
Private, deterministic expense review with an encrypted audit trail
PDF, PNG and JPEG originals are encrypted in your private workspace. Up to 10 MiB each, 100 documents / 100 MiB total. Saving does not start extraction or an Accounting run. Cloud archiving and email intake are planned.
Sign in to view saved originals.
Choose individual files or one whole folder (up to 500 entries). Text is capped at 1 MiB per extracted record / 10 MiB per run. Supported locally: receipt images, PDFs, .txt, .md, .text, .eml, .csv, .ofx, .qfx, and .qif.
Tax-reduction estimates stay blank unless you supply a rate.
Metadata only. Open a run to decrypt its result for this signed-in account.
Loading runs…
Plan campaigns, draft channel content, and keep every revision reviewable.
Turn your business brief into a reviewed plan, channel drafts, or a learning report.
Review workflow opens editable business inputs and the steps before you start. Results stay in Workflows, where you can copy or download them for your campaign. Nothing is published, sent, or scheduled here.
Private client engagements, revision-safe SOWs, milestones, and internal deliverables.
Loading clients…
Loading engagements…
Organize, track, and manage your outreach by campaign
Turn your business brief into a reviewed plan, channel drafts, or a learning report.
Review workflow opens editable business inputs and the steps before you start. Results stay in Workflows, where you can copy or download them for your campaign. Nothing is published, sent, or scheduled here.
| Company ↕ | Contact ↕ | Email ↕ | Financials | Pitch | Subject | Status ↕ | Actions |
|---|
Prospect responses, classified by intent. Hot leads are highlighted green.
Every lead across all campaigns
Your revenue pipeline — forecast by stage and win probability
Your follow-ups — overdue, due today, and upcoming
Bill your clients — draft, send, track paid/overdue, and convert deals into invoices
Admin-only moderation view of completed non-private sessions. Private sessions never appear here.
One quiet place for delegated work and the decisions that genuinely need you.
Setup and changes are not available from this browser view.
Your private goal lifecycle, without prompts or outputs.
Safe readiness summaries only. Account handles and exact usage stay hidden.
Only code-owned profiles with a current policy grant can appear here.
Integrate Meta Council into your applications, agents, and workflows
Go to Settings and create a key in the Developer API Keys section. Keys start with mc_.
curl -X POST https://meta-council.com/api/v1/councils/run \
-H "Authorization: Bearer YOUR_API_KEY" \
-H "Content-Type: application/json" \
-d '{"query": "Should we expand into European markets?", "panel": "default"}'Returns: {"session_id": "abc123", "status": "running"}
curl https://meta-council.com/api/v1/councils/sessions/abc123 \
-H "Authorization: Bearer YOUR_API_KEY"Returns full synthesis, agent reports, risk matrix, and action plan when complete.
| Method | Endpoint | Description |
|---|---|---|
| POST | /api/v1/councils/run | Start a new council session |
| GET | /api/v1/councils/sessions/{id} | Get session status and results |
| GET | /api/v1/agents | List all available agents |
| GET | /api/v1/panels | List all available panels |
| POST | /api/v1/keys | Create a new API key |
| DELETE | /api/v1/keys/{id} | Revoke an API key |
Let your AI agents consult the council before making high-stakes decisions. Get multi-perspective analysis programmatically.
Build a bot that runs council queries from chat. Post the synthesis back to the channel for team discussion.
Run architecture reviews or security audits as part of your deployment pipeline. Gate releases on council confidence scores.
Connect via webhook. Configure your webhook URL in Settings to get notified when councils complete.
Configure your account, models, and integrations
Have a promo code? Enter it above to unlock Pro features.
Get 30 days of free Pro for each person who signs up for Pro using your link.
Choose which model powers your council agents. The platform default (Qwen 3 32B) is free and runs on our servers. External models require your own API key.
| Model | Speed | Quality | $/1M tokens (in / out) | Context | Key Req. |
|---|---|---|---|---|---|
| Qwen 3 32B (default) | ●●●○ | ●●●○ | Free | 128K | No |
| Claude Sonnet 4.6 | ●●●○ | ●●●● | $3 / $15 | 1M | Yes |
| Claude Fable 5 | ●●○○ | ●●●● | $10 / $50 | 1M | Yes |
| GPT-5.6 Sol | ●●○○ | ●●●● | $5 / $30 | 1M | Yes |
| Kimi K3 | ●●○○ | ●●●● | $3 / $15 | 1M | Yes |
| Gemini 3 Flash | ●●●● | ●●●○ | $0.50 / $2 | 1M | Yes |
| DeepSeek V3.2 | ●●●○ | ●●●● | $0.27 / $1.10 | 128K | Yes |
| Groq Llama 3.3 70B | ●●●● | ●●●○ | $0.59 / $0.79 | 128K | Yes |
| Perplexity Sonar Pro | ●●○○ | ●●●○ | $3 / $15 | 200K | Yes |
Add your own API keys to use external models. Keys are stored securely and never displayed after saving. Leave blank to keep the existing key.
Create API keys to integrate Meta Council into your applications. Keys use the mc_ prefix. View API Docs
Save this key now — it won't be shown again!
Connect your Google account to automatically process emails and prepare meeting briefings through workflows.
✅ Connected:
Google connection status is unavailable. Nothing was changed.
Configure account keys consumed by supported built-in data tools. Saved keys are encrypted and never shown again. “Saved” means the credential is on file, not that the external service has been live-verified.
Loading supported tool credentials…
Store a credential only when a supported custom tool or integration documents its exact key name. Saving a name alone does not make it available to agents.
Connect Meta Council to Zapier, Make, n8n, or any webhook-compatible automation platform.
We'll POST a JSON payload here when your council sessions complete.
This protection applies on every tier. Publish a specific non-confidential session explicitly only when you want a public link and search listing.
Connect to your own vLLM, Ollama, SGLang, or LM Studio server. Must expose an OpenAI-compatible /v1/chat/completions endpoint.
Guest runs are public. Your full question, synthesis, and agent reports can be viewed without signing in and may appear in search results.
Sign in for a private run, or explicitly continue with a public run.
Describe a goal and the planner model drafts a ticket tree — steps with substeps, validation and testing included. Preview first; nothing is created until you accept.