Skip to main content

Release 5.0 2025.09.09.1

· 2 min read

🎉 New Features​

Token Distribution in Instances​

Get a detailed view of how your agent is consuming tokens with the new Input Token Distribution charts. These charts show how tokens are allocated across system definitions, knowledge, user messages, skills, and more—helping you understand performance at every step of the conversation.

Token Distribution

MCP Skill​

The new MCP Skill lets your agent connect to remote MCP servers. Simply add the skill and the server URL to start accessing available methods.

MCP Skill

Currently, MCP Skill works only with open servers (No Auth). Support for OAuth is coming soon.

AI Workflows​

Easily build workflows with the help of AI. Interact with Serena to define your needs, and we'll auto-generate a workflow using your available agents. Reduce setup time and get a working workflow faster.

AI Workflows

Support Center​

Find all support resources in one place—FAQs, helpful guides, and contact information—organized for quick and easy access.

Support Center

Provider & Model Control (Compliance Center)​

Gain full control over which AI models are available in your organization. Define allowed providers and models based on compliance or performance requirements, ensuring consistency across your teams.

Provider Model Control

🛠️ Enhancements​

Bring Your Own Key (Compliance Center)​

Bring Your Own Key (BYOK) is now fully integrated into the Compliance Center, unifying all compliance features in a single view. The UI has also been updated for easier navigation.

BYOK Integration

Knowledge​

The Knowledge tab has been refreshed for a clearer and easier presentation of information, improving the overall user experience.

Knowledge - New list of files

Knowledge - New sections list

System Context Parameters​

Quickly see which context parameters are available for your agents.

System Context Parameters

🐞 Bug Fixes & Minor Improvements​

  • Fixed an issue that disabled channels when publishing agents through knowledge API loads, version grid publishing, or knowledge reprocessing.
  • Resolved errors with side panels when duplicating agents multiple times.
  • Fixed issues with sensitive data obfuscation.
  • Simplified conversation methods for improved performance.
  • Recent agents are now automatically updated when a user is updated.
  • Fixed inconsistent disclaimer displays in channels.
  • Added mapping of system context parameters in skills for use in Liquid expressions.
  • Increased initial bot message character limit to 2000.
  • Skill output content is now returned on output tasks in stream mode.