ElevenLabs AI Updates: November 8, 2026
Fully automated. Found, summarized and published by an automated AI pipeline. How posts are labeled
1. ElevenLabs SDKs 2.77.0 Add a Workflow Builder for Defining ElevenAgents Workflows as Code
ElevenLabs released version 2.77.0 of its JavaScript and Python SDKs on November 6 with a Workflow builder for ElevenAgents agent workflows, which previously had to be written through the API as a raw graph of nodes and edges with their own edge IDs, edgeOrder lists and nested expression trees (the pull request notes that anything non-trivial was therefore built in the visual editor). Workflows are now chained from four verbs, converse (a subagent node with its own prompt), tool, transfer (to a phone number, SIP URI or another agent) and end, and routed with when (a plain string is an LLM-evaluated condition, while an Expr such as Expr.variable("tier").eq("gold") tests dynamic variables), onResult, onSuccess and onFailure for tool outcomes, and goTo for loops; build() returns the model that the existing agents.create and agents.update calls already accept. Workflow.fromAgent() loads a deployed agent’s workflow with its node IDs, editor positions and edge order intact for edits through node().update, insertBetween and remove, and validation runs only in build(), which raises one WorkflowValidationError listing every problem (duplicate IDs, unknown goTo targets, tool nodes missing success or failure routes) instead of failing on the first. toJSONString() writes sorted snake_case API JSON in the same shape the CLI uses for agent configs, so a workflow can live in git and be applied from either SDK, the CLI or curl; the two SDKs are tested against shared fixtures and emit byte-identical JSON, and in Python the builder is imported from elevenlabs.agent_workflows with snake_case names such as go_to and from_agent. Source Python release PR