Skip to main content

🧠 Beak AI Studio – Release Notes

Changelog for agents, skills, prompts, datasources, MCP servers, and workspace configuration.

Note: Entries for version 6.0.0 onward are recorded on this page. Earlier releases keep their own detailed pages, listed at the end of this page.

Version 6.0.0​

A new tenant arrives with Studio already set up​

Studio now sets itself up when a tenant is created — its schema, the shared agent tables, the built-in skill types and connector catalogue, and empty configuration documents ready to fill in.

  • The setup checklist says what is left. A Gateway connection is required, because without it the model list is empty and nothing can be tested; knowledge storage and default credentials are optional.
  • Shared knowledge storage is now a choice, not an inheritance. A new tenant is asked for its own storage account, or is told explicitly when it is using the platform's.
  • A Gateway key that cannot be decrypted is now refused rather than passed on. It used to be sent to the Gateway as-is, which came back as an empty model list with nothing in the logs to explain it.

On by default for newly created tenants; existing tenants are untouched. See What Each Product Sets Up.

A failed shared-schema update now stops Studio from starting​

Beak AI Studio already updated the shared agent schema in every active tenant's database at startup. Until now, if that update failed for a tenant, Studio wrote the failure to its log and started anyway, so that tenant could be served against tables that were behind the running code. A failure now stops the new version from starting; the previous version keeps serving until the problem is fixed.

On by default; nothing to configure, and no change when every tenant's schema is already current.

See Platform Administration release notes.

Unified-login sessions no longer expire after 15 minutes​

If you reach Beak AI Studio through unified login, your session now lasts as long as your single sign-on session allows — up to 8 hours on a standard tenant — instead of returning you to the login page roughly every 15 minutes. The renewal request was previously rejected by the browser on tenant-branded addresses — your own company domain, or a <tenant>.beakops.com address — so only users on the default product address were unaffected.

On by default; no configuration, and no change to your tenant's session or idle timeouts. A password sign-in on Beak AI Studio's own login page is unaffected.

See Platform Administration release notes.

Import MCP servers from the Beak Hub marketplace​

MCP servers published to your Beak Hub can now be added from inside Studio. MCP Servers → Import from marketplace lists what your Hub offers, and picking one creates the Hub connection and registers the server in a single step. See MCP Servers for the full flow.

Previously this took three places to finish one task: you chose Beak Hub MCP, pasted a gateway URL you had to already know, then followed a link out to Beak Hub to create the connection before returning to Studio. That URL is no longer typed anywhere, and Studio no longer sends you to Hub to authorise the server.

  • External servers — unchanged. Adding a self-hosted or non-catalogue server is now labelled Add external server, with the same URL, transport, and credential fields.
  • Connection scope — you choose who can use the server, and the default is everyone in the tenant. An only me connection resolves for your agent runs alone, so another user's agent would get no tools from it; pick that only when the credentials are personal.
  • Credentials — passed straight to Beak Hub and stored encrypted there. Studio keeps no copy, the fields are masked, and anything typed into an import you abandon is discarded.
  • Duplicate connection names — refused, with the collision named so you can choose another. Each import creates its own connection, so give them names you can tell apart: Default, Sandbox, Prod.
  • A failed import — cleans up after itself. If the server cannot be registered, the connection just created in Hub is removed, so a failed attempt does not leave a stray connection holding your credentials. If that cleanup cannot complete, the message names the connection to remove.

On upgrade: existing MCP servers keep working and need no action. Servers already pointing at a Hub gateway URL continue to run; their URL becomes read-only, because it is what their Hub connection authenticates.

📄 Available Release Notes​

  • v2.0.0 – Transformational Release:

    • Centralized credential vault and enterprise-grade security controls
    • Generate/download multiple file types; powerful contextual uploads
    • Custom script execution (Python, Bash, PowerShell)
    • Flexible semantic, hybrid, and structured search
    • Complete agent, skill, data source lifecycle management
    • Major integrations (Jira, Teams, webhooks, Copilot interface)
    • Dynamic visualization, conversation summaries, and more

  • v1.1.0 – Major Feature Expansion:

    • MCP Server Support for skill orchestration and agent integration
    • Flexible Uploaded File Datasource (vector DB optional)
    • Expanded Web Datasource Capabilities (websites, portals, SharePoint)
    • Multi-Agent and Architecture Support (orchestration, swarm, collectives)
    • Ambient Agents with Triggers (scheduled/event-based automation)
    • New Agent Architectures: Plan and Execute, Reflexion, Research, Codeact
    • Beak AI Gateway Integration (100+ LLMs, routing, load balancing, failover)
    • Advanced Security Integration (Content Safety, Beak AI Security)
    • Observability and Monitoring (tracing, evaluation, dashboards)
    • Comprehensive File Type Support (documents, spreadsheets, images, etc.)
    • Context Retrieval Enhancements (improved extraction, inline citations)
    • Multi-File Upload (batch onboarding)
    • Chat Widget Deployment (embed in web platforms)
    • Model Controls in Playground (model selection, temperature)
    • Custom Skill Creation (REST API, LLM prompt, Python code skills)
    • Role-Based Access Control (RBAC) for agents, datasources, skills
    • Automated Deployment with Terraform
    • Pre-Built AI Agents (CloudOps, ITOps, Service Desk, Knowledge, and more)

  • v1.0.0 – Initial Release:

    • Custom agent creation and skill support
    • Prompt playground and data source integration
    • Agent publishing, web chat deployment, observability and analytics
    • CI/CD-ready deployment automation with Terraform