Skip to content
NewsProduct launch

n8n 2.34.0 ships workflow review gates, per-agent MCP scopes, and a task runner fix

· by Pondero Newsdesk

The short version

n8n released version 2.34.0 on August 4, adding a formal workflow review and approval system, per-agent MCP access controls, and OAuth authentication for webhook triggers. Four patch releases through August 7 fixed a task runner health check failure in scaling deployments.

n8n 2.34.0 ships workflow review gates, per-agent MCP scopes, and a task runner fix

Before August 4, any n8n user with publish access could push a workflow live in a single click. Version 2.34.0 changes that: a formal review cycle now sits between a workflow change and its production deployment, with a reviewer able to approve or request changes, the system blocking publishing while a review is open, and the workflow auto-publishing once approved. The same release ships per-agent MCP scope controls and OAuth authentication for webhook triggers. Four patches over the following three days fixed a task runner health check failure in scaling deployments.

What changed

The workflow review system is the most structurally new capability per the 2.34.0 release notes. Reviewers can approve a workflow or request changes via a new detail endpoint. A guard blocks publishing any workflow while its review is open. An approved review triggers automatic publication without a separate manual step.

The editor gained per-agent MCP access management. Each agent now has its own set of MCP scopes rather than sharing a flat access model across all agents on an instance. Combined with the existing MCP server usage events in log streaming (also in this release), operators running multiple agents from a single n8n instance can now both restrict what each agent can reach and audit what it accessed.

A new n8n user auth (OAuth2) mode arrived for webhook triggers. A "Require Workflow Execute Permission" toggle gates which authenticated users can fire a given webhook-based workflow. OIDC logout support was added at the same time.

For scaling mode, large webhook responses now offload to storage rather than flowing through the broker, cutting memory pressure on main-process instances. Self-hosted setup got a simpler entry point: a one-script Docker Compose installer is now available at get.n8n.io.

Four point releases followed between August 5 and 7. The recurring fix was a task runner health check failure that caused health checks to report false negatives in queue-mode configurations with external task runners, resolved in 2.34.4 and its stable-branch equivalent 2.33.7 per GitHub. A pre-existing production SSE push backend connection issue was also addressed in 2.34.0 itself; it had been silently breaking real-time execution feedback in some deployments.

Why it matters

The workflow review gate matters most to teams where AI automations touch anything with side effects: outbound emails, CRM writes, Slack messages, external API calls. Without it, a bad workflow publish was a silent risk. The approval step gives a designated reviewer a checkpoint before changes reach production, without requiring custom deployment tooling outside n8n.

Per-agent MCP scopes close a governance gap that grew as operators started running multiple agents from a single n8n instance. One compromised or misbehaving agent could previously reach every MCP tool available on the instance. Scoped access limits the blast radius.

Self-hosted operators running queue mode should review the 2.34.4 release notes before deploying; a false-negative health check on a task runner can cause jobs to queue indefinitely rather than fail fast with a diagnostic error.

What to watch next

The current workflow review system is a single-layer approval loop. A multi-stage chain (developer proposes, team lead approves, admin publishes) would be a natural extension for larger teams. The public API endpoint for review requests opened in this release gives enough surface area to build that externally now; watch whether n8n adds native multi-stage support in a future minor release. Per-agent MCP scope documentation, currently thin, is the other gap operators will push on once adoption grows.

Sources