Why is the Agent node (v2) unavailable in Chatflow but available in Workflow? Any plans to support it? #42247

The restriction is intentional and comes from the frontend, as you suspected. Agent V2 was hidden from Chatflow in v1.16.1 via PR #39439 ("fix(workflow): hide Agent V2 in chatflow"); in v1.16.0 it was still selectable there.

The gate lives in useAvailableNodesMetaData(), which computes shouldUseAgentV2 = agentV2Enabled && (!isChatMode || isAgentV2InChatflowEnabled()). Since NEXT_PUBLIC_ENABLE_AGENT_V2_IN_CHATFLOW defaults to false, the node picker filters Agent V2 out whenever isChatMode is true. The backend handles both versions and resolves conversation_id in chatflow runs, which matches what you found in the code.

The reason for hiding it: a maintainer explained in #41269 that Agent V2 didn't preserve conversation-level context correctly for multi-turn chats. Work has since landed to key Agent V2 sessions by conversation_id instead of workflow_run_id when running in Chatflow, enabling multi-turn memory persistence (PR #41871).

On plans, the same maintainer confirmed in #41269 that official Chatflow support is planned with multi-turn behavior consistent with the standalone Agent experience, though no target version is confirmed yet.

As a workaround on self-hosted deployments, set NEXT_PUBLIC_ENABLE_AGENT_V2_IN_CHATFLOW=true to opt into the in-progress Chatflow support ahead of the official rollout.

Want me to dig into the current state of the multi-turn memory work to gauge how close it is to being enabled by default?

To reply, just mention @dosu.

Docs are dead. Just use Dosu.

Agent v2 is currently in beta and is only enabled in Workflow mode. The team has it gated behind a feature flag (NEXT_PUBLIC_ENABLE_AGENT_V2_IN_CHATFLOW) which is false by default. You can enable it in your web/.env.local by setting NEXT_PUBLIC_ENABLE_AGENT_V2_IN_CHATFLOW=true, but keep in mind it's still experimental in Chatflow. The reason it's not exposed yet is that the human-in-the-loop (ask_human) flow in Chatflow requires conversation context that Agent v2 doesn't fully support yet. For production use, I'd recommend sticking with Workflow mode for Agent v2 for now.