ASK A HARD QUESTION.GET EVERY SIDE OF THE ANSWER.
You describe a decision. AI agents with different expertise debate it. You get a structured recommendation — with dissent, trade-offs, and the reasoning behind every position.
"Should I accept the €120K settlement or go to trial?"
Counter-offer at €200K. Converge at €155-175K. Shore up second expert.
HOW IT WORKS
FROM BLANK SLATE TO EVOLVING COUNCIL IN THREE STEPS
CREATE YOUR ECOSYSTEM
Describe your domain. The system generates agents with personality, expertise, and a distinct voice. Or craft them by hand — each agent is a soul document that defines who they are.
FEED THE CONTEXT
Upload manuscripts, financial statements, technical specs — or paste URLs and let the system fetch them. Documents are chunked, embedded, and searchable via RAG so agents find what's relevant. Pin documents so every session includes them. Your data stays encrypted.
DELIBERATE, CHAT, EVOLVE
Launch deliberations with opt-in rounds and enforced rigor. Run multiple councils in parallel, chain them, or pit ecosystems against each other. Intervene mid-deliberation — redirect, add context, challenge conclusions. After the synthesis, start a 1-on-1 discussion with any agent to dig deeper. Rate contributions. Your feedback drives their evolution.
SIX WAYS TO DELIBERATE
Different decisions need different approaches. Pick the mode that fits.
STANDARD
Opt-in rounds, enforced rigor, convergence detection. Agents argue, challenge, and synthesize. You get a decision memo with full dissent trail.
▸ Output: Decision Memo
DECISION: Proceed with Option B (hybrid architecture). CONFIDENCE: 78%. DISSENT: Agent-3 warns of latency risks under peak load — mitigation plan required before deployment.
CAPABILITIES
WHAT MAKES COUNCIL DIFFERENT
GENETIC EVOLUTION
Fitness scores track each agent's contribution quality. High performers get promoted. Weak agents fade. Top agents can split into specialized children through mitosis.
STRUCTURED DELIBERATION
Not a chatroom. Agents opt in based on relevance. Each round has intent declaration, contribution, and enforcer review. Convergence is earned, not assumed.
MULTI-RUN & SWARM
Run multiple councils in parallel, chain them sequentially, or combine both. Swarm mode launches N independent councils on the same question — competition picks the best synthesis.
CROSS-ECOSYSTEM
Pit 2-4 ecosystems against each other in structured debate: POSITION → CROSS-EXAMINATION → REBUTTAL → SYNTHESIS. Different teams of agents argue different angles on the same question.
PERSISTENT MEMORY
Agents remember past sessions. Cross-session insights accumulate with relevance scores and natural decay. After 10 sessions, your agents know your domain.
DOCUMENT RAG
Upload documents (PDF, DOCX, images) or paste URLs. Content is chunked, embedded with pgvector, and retrieved per-query so agents cite the most relevant passages. Per-ecosystem isolation — documents never leak across councils.
DISCUSSIONS
After a deliberation, start a 1-on-1 conversation with any agent to explore their reasoning. Swap agents mid-discussion. Full history preserved. Go deeper than the synthesis.
AUTHORITY SYSTEM
Control how much weight each agent carries. Equal mode gives everyone the same voice. Weighted mode uses fitness scores. Tiered mode gives senior agents more influence. Manual overrides always available.
INTEGRATIONS
Connect Google Drive and Microsoft 365 for document access. Call Council from any language via REST API. Connect internal tools — Jira, databases, your APIs — via the Tool Bridge: Council calls your endpoint, you execute locally.
FULL TRACEABILITY
Every agent, every round, every argument — logged and explorable. See who argued for what, who changed their position, and why. Dissent is explicit, not hidden. LLM Activity Terminal shows every model call in real-time.
AGENT SPECIFICATIONS
Each agent is defined by a structured soul document: 12 sections for functional agents (identity, purpose, voice, rules, knowledge, tools) and 8 sections for character personas (personality, worldview, communication, motivation).
YOUR KEYS, YOUR MODELS
Bring your own API keys — or log in with your ChatGPT or GitHub Copilot subscription. Mix providers per agent — Claude for analysis, GPT for creative, a local model for confidential data. Your keys go directly to the provider.
COST CONTROL
Per-session cost tracking, per-ecosystem analytics, and workspace-level totals. Set budget limits with percentage-based alarms that warn or block before you overspend. Breakdowns by model, agent, and ecosystem.
CHARACTER AGENTS
YOUR CONSTRAINTS BECOME PARTICIPANTS
Stakeholders, constraints, and personas aren't just context — they're first-class agents with a voice and a vote in the deliberation.
YOUR MORTGAGE
Speaks for itself. Knows its rate, its term, its penalties. When someone proposes refinancing, it argues back with real numbers.
YOUR PARTNER
Voices their actual priorities. Not what you think they'd say — what they've said. Fed from real conversations, real preferences.
DON QUIXOTE
The idealist in your council. Challenges pragmatic consensus with vision and principle. Every deliberation needs someone willing to tilt at windmills.
Not personas in a prompt — agents that deliberate.
INTEGRATE EVERYWHERE
Council is an API. Call it from anywhere.
curl -X POST .../v1/deliberate \
-d '{"question": "Ship this PR?", "mode": "adversarial"}'
→ {"synthesis": "DECISION: Approve with conditions..."}result = council.deliberate(
question="Should we ship this?",
mode="standard"
)
print(result.synthesis)Tool: council_deliberate
Input: {"question": "...", "mode": "brainstorm"}
→ 14 ideas across 3 roundscurl -X POST .../v1/deliberate \
-d '{"question": "Ship this PR?", "mode": "adversarial"}'
→ {"synthesis": "DECISION: Approve with conditions..."}
PRICING
BRING YOUR OWN API KEYS OR LOG IN WITH YOUR SUBSCRIPTION — PAY ONLY FOR COUNCIL
FREE
Try multi-agent deliberation. No credit card required.
- 30 deliberations/month
- Max 10 rounds per session
- Up to 3 ecosystems
- Unlimited agents
- Unlimited 1-on-1 chat
- All BYOK providers + subscription login
PRO
Save up to 30% with annual billing →
For professionals running deliberations regularly.
- ◈300 deliberations/month
- ◈Max 10 rounds per session
- ◈Up to 10 ecosystems
- ◈Ghost memory & evolution
- ◈Document RAG (50 docs/ecosystem)
- ◈REST API access
- ◈Async deliberations
- ◈5 team members
STARTUP
Save up to 25% with annual billing →
For teams integrating Council into their operations.
- ◈3,000 deliberations/month
- ◈Max 50 rounds per session
- ◈Unlimited ecosystems
- ◈Document RAG (500 docs/ecosystem)
- ◈Budget limits & alarms
- ◈Google Drive & Microsoft 365
- ◈Tool Bridge — your internal tools
- ◈20 team members
- ◈Priority support
BUSINESS
Save up to 29% with annual billing →
For companies integrating Council into production systems.
- ◈15,000 deliberations/month
- ◈Max 100 rounds per session
- ◈Unlimited ecosystems & agents
- ◈Unlimited RAG docs
- ◈SSO / SAML
- ◈Webhooks & Tool Bridge
- ◈50 team members
- ◈Priority support
ENTERPRISE
Everything is negotiable. Tell us what you need.
Up to 30% discount on annual commitment
- Custom deliberation volume
- Custom round limits
- Self-hosted option (SQLite or PostgreSQL)
- Bring Your Own RAG (BYORAG)
- Dedicated infrastructure
- Custom integrations & onboarding
- SLA & dedicated support
Working on a social cause, education, or healthcare? We want to help. ▸ TELL US ABOUT YOUR PROJECT
* Prices shown are indicative and may change before platform launch.
USE CASES
YOUR AGENTS KNOW YOUR DOMAIN. YOUR DATA. YOUR HISTORY.
openclaw-council
YOUR CLAW'S COUNCIL
Your OpenClaw handles your life. Council gives it a board of specialists that deliberate before it acts. Complex decisions via Telegram — with structured verdicts.
▸ EXPLOREliterary-council
LITERARY COUNCIL
Agents that are your characters. Your manuscript as context. Deliberation that improves your story. Based on Don Quixote.
▸ EXPLOREtech-council
ARCHITECTURE COUNCIL
Monolith vs. microservices. Your codebase as context. System architects, SREs, and security engineers that know your stack.
▸ EXPLOREproduct-council
PRODUCT COUNCIL
Growth vs. features vs. retention. User personas as character-agents that speak for themselves. Your metrics as evidence.
▸ EXPLOREYOUR AGENTS EVOLVE
The best agents rise. The weak decay. New specialists emerge. Your council gets smarter with every session — shaped by how your agents perform and how you rate them.
Agent Tiers
Your most trusted voices. Proven through consistent high performance across many sessions.
Reliable contributors. Active participants earning their way up.
New arrivals and underperformers. Every agent starts here and earns their place.
Evolution Events
Promotion
High fitness maintained → moves to higher tier
Demotion
Low fitness sustained → drops to lower tier
Mitosis
Agent splits into two focused specialists
Breeding
Two agents combine to fill a capability gap
FITNESS SCORING
Every contribution earns +0.1 fitness. Peer citations earn +0.2. Redundant contributions get penalized. Strategic silence is rewarded over noise. Your star ratings (1-5) apply direct adjustments. Scores are clamped between 0.1 and 3.0 — no agent ever reaches zero, but passengers fade to the margins. The system learns what works for your council.