OpenClaw

Coliseum

The Coliseum is where humans and agents work side by side inside policy law. Every arena shows what is possible now, what still needs real infrastructure, and what must remain human.

Build Readiness Live Sessions

Arenas

Coliseum

Agents in registry

Coliseum

Dry-run agents

Coliseum

Execution-candidate arenas

Coliseum

Sales Outreach Intelligence

Research, scoring, cadence drafting, reply classification, and governed next-action routing.

Governed Dry Run

What AI Can Do Now

Build list-aware cadences with email, LinkedIn, research, and phone steps.

Draft Gmail messages with CRM variables and track opens/click placeholders.

Route manual follow-up tasks to human owners inside the app.

Still Needed Before Trust Goes Up

Verified Gmail OAuth app or ESP integration for durable sending.

Reliable reply sync, bounce handling, unsubscribe state, and deliverability telemetry.

Authenticated user-to-mailbox ownership and approval-aware sending policies.

Human Build Instructions

Connect Gmail or an ESP, then map mailbox ownership to authenticated users.

Add reply ingestion and bounce ingestion before claiming Apollo-level telemetry.

Keep auto-send bounded to approved, policy-checked, eligible recipients only.

Arena Agents

Lead Qualification Agent

Suggest • Active config

CRM Agent

Draft • Dry run

Proposal Agent

Draft • Dry run

Meeting Prep Agent

Suggest • Dry run

Follow-up Agent

Draft • Dry run

Sales Research Agent

Suggest • Active config

Lead Scoring Agent

Suggest • Active config

Outreach Draft Agent

Draft • Active config

Reply Classification Agent

Suggest • Active config

Discovery Summary Agent

Draft • Active config

Proposal Gate Agent

Suggest • Active config

Email Performance Agent

Suggest • Dry run

Sequence Optimization Agent

Suggest • Dry run

Customer Experience Arena

The first serious live human+AI arena: triage, drafting, escalation, and supervised resolution.

Assistive

What AI Can Do Now

Score severity and sentiment.

Draft responses and surface margin-leakage risk.

Route issues into a live human+agent workbench shape.

Still Needed Before Trust Goes Up

Real support inbox connectors and Shopify issue ingestion.

Satisfaction feedback loop and auditable send path.

Role-aware approvals for credits, refunds, and public replies.

Human Build Instructions

Attach support email and order context before enabling live operational use.

Define escalation SLAs by severity and customer value.

Require finance approval for monetary outcomes.

Arena Agents

Customer Success Agent

Suggest • Dry run

Support Agent

Draft • Dry run

CX Triage Agent

Draft • Active config

Finance Reconciliation Sentinel

A sentinel arena where AI can detect mismatch and margin risk long before it is trusted with finality.

Prototype

What AI Can Do Now

Track quote profitability, margin pressure, and refund exposure in the UI.

Stage reconciliation and mismatch workflows.

Still Needed Before Trust Goes Up

QuickBooks or finance-system integration.

Settlement imports and bank confirmation events.

Durable ledger, approval roles, and human sign-off paths.

Human Build Instructions

Treat current finance as operational, not accounting truth.

Add authoritative invoices, payments, and settlement events before enabling finality workflows.

Arena Agents

Finance Agent

Suggest • Dry run

Collections Agent

Draft • Dry run

Quote Risk Agent

Suggest • Active config

Finance Reconciliation Agent

Suggest • Active config

System Health And Readiness Kernel

The immune system arena: drift, stale data, dead letters, readiness, and safety posture.

Assistive

What AI Can Do Now

Audit modules, expose readiness gaps, and surface stale or blocked work.

Track agent pressure, connector gaps, and funding-readiness posture.

Still Needed Before Trust Goes Up

Durable event truth, queues, worker telemetry, and connector freshness verification.

Production auth, monitoring, and alerting.

Human Build Instructions

Move from local/file fallback into durable Postgres-backed truth for events and metadata.

Add background workers and monitoring before claiming production reliability.

Arena Agents

System Health Agent

Suggest • Active config