Encord Launches Merlin, Agentic Intelligence Layer for AI Data Infrastructure
Key Takeaways
- ▸Merlin is the first natively embedded agentic intelligence layer in an AI data infrastructure platform, allowing users to manage data workflows through conversation
- ▸Supports the complete data lifecycle: building label schemas and review workflows from prompts, monitoring real-time data metrics, and optimizing model performance by identifying data issues
- ▸Multi-platform launch strategy includes MCP, Claude, Codex, and other agentic platforms, with Slack integrations coming soon
Summary
Encord has introduced Merlin, an agentic intelligence layer that embeds AI agents directly into its data infrastructure platform to automate the full data lifecycle for machine learning projects. Merlin works across build, observe, and optimize stages—helping teams define label schemas, set up review workflows, monitor data metrics and coverage gaps, and identify performance bottlenecks—all through conversational, agent-driven interfaces. The platform is launching in beta via Model Context Protocol (MCP) and integrates with Claude, Codex, and other agentic coding platforms, with Slack and additional integrations coming soon. This addresses a critical pain point in AI development: the manual, time-consuming work of preparing raw data for training.
- Targets a major bottleneck in ML development: manual data preparation, reducing weeks of manual work to instant project setup
- Early access program open to selected customers, with plans to expand agentic capabilities across Encord's platform
Editorial Opinion
Embedding agents into domain-specific tools like Encord represents a powerful emerging pattern that could significantly accelerate ML development workflows. Data labeling and preparation are notoriously tedious bottlenecks in ML projects, and intelligent agents that can handle these tasks conversationally could save teams weeks of manual work. Encord's multi-platform approach—launching across Claude, Codex, and MCP simultaneously—signals a smart bet on the agent-first future of development tools.
