Ai

Agno v2.6.0 Adds Human Oversight, Multi-Framework AgentOS Support and Runtime Context Tools

Agno v2.6.0 Adds Human Oversight, Multi-Framework AgentOS Support and Runtime Context Tools

Agno v2.6.0 is a meaningful platform update focused on making agent systems more governable, more interoperable, and easier to run in production. This release adds human-in-the-loop controls across teams and workflow execution, introduces approvals directly in Team and AgentOS, expands AgentOS toward multi-framework compatibility, and improves operational resilience with reconnection and resume support for background runs.

What Changed

The headline change in v2.6.0 is expanded human oversight. Agno now includes an API layer for Team human-in-the-loop workflows, plus corresponding support in the AgentOS chat page. Team approvals were also added, making it easier to insert explicit sign-off steps into collaborative agent flows and surface them directly in AgentOS.

The workflow layer also got a human-in-the-loop upgrade. Agno now supports executor-level intervention when a pause tool flow is configured on an agent or team within a workflow step. That gives operators a more granular checkpoint mechanism during multi-step execution rather than only at the top-level team layer.

AgentOS itself is becoming more of a unifying runtime. In beta, v2.6.0 adds basic multi-framework support for ClaudeAgentSDK, Langgraph, and DSPy through a shared AgentProtocol backbone. That matters because it positions AgentOS as a common operational surface for heterogeneous agent stacks instead of only one native framework path.

Another practical improvement is run continuity. Agents and teams running in the background over SSE can now reconnect and resume in AgentOS after interruptions such as a page refresh. For longer-running tasks, that reduces operational friction and lowers the risk of losing visibility into active runs.

The release also introduces dynamic runtime creation through AgentFactory, TeamFactory, and WorkflowFactory. These new factory components are aimed at multi-tenant and programmatic deployment scenarios where agents, teams, and workflows need to be instantiated on demand rather than defined statically ahead of time.

Finally, Agno added agno.context, a first-party context provider API. It lets developers connect external sources such as filesystems, websites, SQL databases, Slack, Google Drive, and MCP servers into an agent as natural-language tools. That broadens how agents can access enterprise knowledge and operational data without forcing each integration into a custom pattern.

Why It Matters

For teams building production-grade agent systems, v2.6.0 strengthens three areas that matter most: governance, interoperability, and runtime reliability. Human-in-the-loop and approval support make Agno more suitable for workflows where autonomy must be balanced with review and control. That is particularly relevant in enterprise environments, regulated processes, and customer-facing automations.

The multi-framework AgentOS direction is also strategically important. By supporting ClaudeAgentSDK, Langgraph, and DSPy through a common protocol layer, Agno is moving toward a more open control plane model. That can help engineering teams avoid framework lock-in while keeping a consistent runtime and chat experience.

The addition of resumable background runs and dynamic factories further improves Agno's appeal for multi-tenant SaaS and operational deployments where long-running jobs, interruptions, and on-demand provisioning are normal. Combined with the new context provider API, the platform is clearly pushing toward more flexible real-world enterprise agent orchestration.

Bug Fixes Included

Alongside the feature work, v2.6.0 fixes callable factory behavior so factory tools and members are preserved when Agent or Team objects are copied, and avoids iterating callable-factory tools during deep copy. The Telegram interface was also improved to respect retry_after values on 429 rate-limit responses, which should make integrations more robust under messaging pressure.

Official Source: https://github.com/agno-agi/agno/releases/tag/v2.6.0