ElevenLabs Launches ElevenAgents Architect to Automate Agent Improvement

ElevenLabs has launched ElevenAgents Architect, a built-in AI expert designed to help teams launch and improve conversational agents through natural language interaction. The tool analyzes agent performance signals, such as failing tests or escalation spikes, to identify root causes and propose configuration changes for knowledge bases, prompts, workflows, and guardrails.

Proactive Improvement and Validation

ElevenAgents Architect reasons across thousands of conversations to uncover improvement opportunities that human teams might miss. It drafts changes and validates them through simulated conversations before presenting them for approval. This allows teams to focus on deciding what to ship rather than hunting for issues or manually building each fix.

Safety and Team Empowerment

All changes are implemented as versioned drafts, requiring explicit human approval before going live. The system adheres to team-defined standards and helps establish guardrails and edge-case simulations during the build process. This approach enables non-specialists, such as customer experience leads and compliance officers, to update agents independently, freeing technical experts for higher-level architecture and integration tasks.

Integration and Availability

Users can interact with ElevenAgents Architect via voice or chat within the ElevenLabs product or through external tools including Claude, Claude Code, ChatGPT, Cursor, and Grok Bot. The tool brings full context about the user's agents and ElevenAgents best practices to these platforms. ElevenAgents Architect is currently available in Alpha.