Agentic AI Platform Change Log

Workflow orchestration

Replit can now find a user's existing projects from a chat, read their files and send tasks to them without opening each project. Changes run as background tasks the user can review, and nothing is published automatically.

Teams running many internal apps can push one change, such as a fix or a branding update, across several apps from one place while keeping a review step before anything goes live.

Security / enterprise

Workspace settings joined Replit's enterprise controls. Account admins set organization wide policies, make exceptions for particular workspaces and hand some settings to workspace admins, while locked controls, such as limiting public deployments to admins, stay fixed.

Large organizations can hold central rules on who may publish apps publicly while letting teams manage their own settings.

Agent capability

Replit Agent can now analyze an uploaded dataset or connected data source and display interactive charts directly in chat, using technology from its acquisition of Atta. Users can request a specific chart or ask the agent to visualize findings, with the agent handling queries and analytical steps.

Business teams can investigate data and inspect visual results within the same conversation, reducing the need to hand each follow up question to a separate analyst or charting tool.

Pricing / packaging

Replit introduced Free Mode, a new tier powered by OpenAI's GPT-5.6 Luna model that allows users to build applications and agents without consuming token usage. For tasks requiring more advanced reasoning, the platform routes requests to GPT-5.6 Sol before returning to the free Luna model while preserving project context.

This removes usage-based token costs for standard development tasks, significantly lowering the barrier to entry. Evaluators can now extensively prototype and test the agent's core capabilities before committing to a paid subscription.

Integrations

Replit Agent now supports migrating eligible Replit Auth applications to Clerk Auth. The agent automatically handles moving account data before updating the sign-in code to ensure returning users retain their access.

This allows teams to transition their authentication infrastructure to a robust third-party provider without losing existing user data or manually rewriting sign-in logic.

Security / enterprise

Replit introduced a suite of updates highlighted by automated security scanning, allowing the Agent to check modified files for vulnerabilities during code review. The release also adds production database credential regeneration, enterprise SSO for Clerk Auth apps, workspace-to-workspace project transfers, and native recognition for the Braintrust MCP endpoint.

Enterprise buyers gain stronger compliance and access controls through automated vulnerability checks, SSO integration, and configurable session timeouts. Developers and teams benefit from smoother collaboration via project transfers and expanded extensibility with Braintrust MCP support.

Agent capability

Replit introduced a continual learning system for Replit Agent, powered by an internal agent-based architecture. The system automatically processes user feedback, proposes system improvements, and validates changes using benchmarks and A/B testing.

Buyers benefit from an agent that continuously updates its coding accuracy and problem-solving logic based on real-world usage. This automated feedback loop enables the product to improve over time without relying entirely on discrete, major version releases.

Agent capability

Workspace-level Agent customization added reusable workspace instructions plus on-demand skills.

Changes how teams standardize agent behavior across projects, improving consistency and repeatability.

Security / enterprise

Package Firewall now blocks malicious or compromised packages by default, and Agent adapts its plans to safer alternatives.

Material governance and security improvement for production coding agents — reduces supply-chain risk in agent-assisted builds.