Configure
Define prompts, guardrails, approved responses, and content in one managed agent lifecycle.
- Prompts, policies, and guardrails versioned together
- Content from help centers, PDFs, and URLs
- Actions connected to your systems
DEVELOPERS
Build managed AI agents.
Configure, test, deploy, and monitor agents for customer support from the API, CLI, or dashboard — as versioned, auditable releases.
Auth
device login
Agent
support-agent
Revision
v42 preview
Mode
managed
$
From first prompt to production agent.
Every agent lifecycle — configure, test, deploy, and monitor — managed through the same versioned release system.
Test
Run simulation suites against your agent before any change reaches live traffic.
- Multi-turn conversation simulations
- Coverage by intent, topic, and policy
- Regression tests after every change
Deploy
Publish versioned releases — CLI, API, or dashboard — with dry-run validation first.
- applied agents deploy from the CLI
- REST API for programmatic releases
- Dry-run validates before production traffic
Monitor
Stream request logs, audit policy decisions, and track KPIs from one operating surface.
- Live tail of conversations, actions, and events
- Policy review queue for blocked actions
- Volume, latency, and KPI dashboards
Everything you need to build production-grade agents.
REVISIONS
Build and deploy agents as managed releases.
Define prompts, guardrails, approved responses, content sources, and live revisions in one agent lifecycle.
- ✓Versioned revisions. Every prompt, policy, and content change becomes an auditable release.
- ✓Branching workflows. Teams can merge prompt, policy, and content branches independently.
- ✓Dry-run validation. Test every release against your simulation suite before deploying.
- ✓Live variants. Run A/B experiments on competing revisions with real traffic.
support-agent
Every prompt, policy, content, and deploy change becomes an auditable agent revision.
Production
prod branch
Live revisions
Variant A + B
Contributors
5 merged changes
OBSERVABILITY
Log and govern every agent action.
Stream request logs, events, actions, policy decisions, and audit context so teams can debug production behavior without losing traceability.
- ✓Live tail. Stream conversations, messages, events, and action results in real time.
- ✓Policy audit. Every guardrail decision is logged with context and rationale.
- ✓Structured events. POST /v1/events from your systems to trigger agent workflows.
- ✓Conversation spans. Inspect the full trace for any resolved or escalated conversation.
Today
Tail every request, event, action, and policy decision.
14:44:20.554reviewPOST/v1/agents/support/revisions
14:44:21.103200 OKGET/v1/content/search
14:44:22.447204 OKPOST/v1/events
14:44:24.009blockedPOST/v1/actions/refund
14:44:25.612200 OKGET/v1/conversation_spans
14:44:27.018202 queuedPOST/v1/simulations
14:44:18.221200 OKPOST/v1/conversations
14:44:19.006202 queuedPOST/v1/messages
14:44:19.884200 OKPOST/v1/messages
14:44:20.118202 queuedPOST/v1/flow_runs
ANALYTICS
Monitor launches with developer analytics.
Track request volume, event delivery, policy reviews, latency, simulations, spans, experiments, and flow runs from one operating surface.
- ✓Request metrics. Volume, latency, and error rates for every API surface.
- ✓Simulation tracking. Monitor test run coverage and regression results over time.
- ✓KPI dashboards. Confidence, automation rate, and CSAT indexed to each revision.
Documentation and references.
Start building your first agent.
Read the docs, browse the API reference, or book a demo to see the full developer surface.