← Back to all sparks
P

Port

INFRA · APIS
Velocity3.8

Internal developer portal for platform engineering, service catalogs, and scorecards.

Port is becoming the governed orchestration hub for AI agents operating on engineering infrastructure.

developer-platformai-governancemcpworkflow-orchestrationagent-managementengineering-intelligence
Current state
Port shipped two landmark releases in the past two months: Port Workflows reached GA as a governed orchestration layer for agents, automations, and approvals — with Cursor Cloud Agent syncing and triggering built in — and Port AI Builder gained the ability to execute real catalog changes (create blueprints, trigger workflows, upsert entities) with a plan-before-execute approval model. The platform now syncs Claude and Cursor managed agents into the catalog, routes approvals through Slack, scopes workflow permissions to MCP invocations, and lets teams define which AI agents can trigger which workflows. The cadence is roughly one major platform expansion per month.
Where it's heading
Port is converging on two roles: the catalog and context layer that AI agents query to understand the software stack, and the orchestration and governance layer that defines what those agents are allowed to do. The Skills framework, the MCP server, the permissions model for MCP-invoked workflows, and the Survey Intelligence tool all reinforce the same thesis — Port wants to be the enterprise control plane for AI coding tools. Every monthly release adds agent integrations (Claude, Cursor, Codex) alongside governance primitives (permissions, approvals, audit logs).
Prediction
The next move is likely extending workflow permissions to cover more agent types (Codex, GitHub Copilot agents) and adding workflow-level observability so platform teams can audit what AI agents execute at the orchestration layer — closing the governance loop that the MCP source permissions started.

Recent moves

  1. 6d ago

    Port Workflows reaches GA with Cursor agent orchestration and MCP permission scoping

    ⚡ SPARK

    Port Workflows reaching GA is the foundational move — it defines the governed orchestration layer that everything else (Cursor agent sync, Slack approvals, MCP permissions) plugs into. Adding Cursor Cloud Agent sync and triggering from workflows means Port can now orchestrate AI agent work alongside human approvals in a single governed flow.

  2. 1mo ago

    Port AI Builder executes real catalog changes; Claude Managed Agents integration ships

    ⚡ SPARK

    Port AI Builder executing real catalog changes — with plan-before-execute and per-tool approval controls — shifts Port from a query interface to an active agent operating on engineering infrastructure. The Claude Managed Agents integration in the same release extends the catalog to include AI agent activity as first-class catalog data.

  3. 2mo ago

    Port Product Updates - What we built in June 2026

    Workflows in open beta with GitLab CI/CD triggers and external MCP server connections expands the orchestration surface significantly — more CI/CD systems covered, more external tools reachable from inside Port AI. The private pages and Slack thread memory round out usability gaps that would slow adoption.

  4. 3mo ago

    Port Product Updates - What we built in May 2026

    The public port-plugins repo opens the extensibility model and is the earliest signal of Port's ecosystem ambition. The multiple workflow triggers and sidebar redesign both reduce friction for teams building complex automation flows, and tracking Claude Code / Copilot usage natively extends Port's AI observability story.

  5. 4mo ago

    Port Product Updates - What we built in April 2026

    Custom Widgets let teams embed their own React/JS components inside Port, and the any-OpenAI-endpoint setting makes Port AI work with self-hosted and Azure-hosted models — two important enterprise flexibility levers. Structured output schemas for AI workflow nodes make agentic automation reliable enough to use in production pipelines.

  6. 5mo ago

    Port Product Updates - What we built in March

    External MCP server connections in Port AI is the first step toward making Port a unified gateway for the whole tool stack — Atlassian, Notion, GitHub, and any other MCP server can be reached through Port AI's side chat. Visual data source mapping removes a YAML friction point that slowed catalog setup for non-technical admins.