LoginSign Up
← Back to all sparks
P

Port

INFRA · APIS
Velocity2.5

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

www.port.io ↗

Port is turning its developer catalog into an AI- and MCP-native control plane.

internal-developer-platformai-agentsmcpextensibilitygovernanceintegrations
◆Current state
Port has spent the last two quarters converting its internal developer platform into an AI-and-agent surface. Nearly every monthly release now leads with Port AI: an MCP gateway, bring-your-own-LLM routing, agent governance, and now an opening plugin ecosystem. The underlying catalog, scorecards, and RBAC work continues, but it increasingly serves as context the AI layer reasons over rather than the headline itself.
◆Where it's heading
The direction is a platform you build on and talk to, not just configure. MCP connectors, custom widgets, a public plugins repo, and structured AI outputs all point to Port positioning itself as the governed entry point for agentic engineering workflows. Governance is keeping pace deliberately — permission simulators, audit logs, and per-trigger access controls ship alongside each AI expansion, which signals an enterprise buyer.
◆Prediction
Expect the plugins repo and custom widgets to converge into a first-class marketplace, and the Claude Code/Copilot usage tracking to grow into broader AI-spend and agent-activity analytics across the catalog.

◆Recent moves

  1. 21d ago

    Public plugins repo, multi-trigger workflows, Claude Code & Copilot tracking

    May continues the extensibility arc: a public port-plugins repo is the first concrete step toward a community ecosystem, while workflows gain multiple triggers and per-trigger permissions. New Claude Code and Copilot usage integrations extend the AI-tooling visibility Port began tracking last year. Substantial but incremental — these widen surfaces already established rather than open a new one.

  2. 1mo ago

    Custom Widgets, any OpenAI-compatible LLM, and structured AI output schemas

    ⚡ SPARK

    April is where Port becomes genuinely extensible: Custom Widgets let teams build and run their own frontend components inside Port dashboards, and any OpenAI-compatible endpoint can now back Port AI. Structured output schemas make AI workflow nodes reliable enough to chain into automations. This is the month the platform stopped being a fixed UI over a catalog.

  3. 2mo ago

    MCP Connectors turn Port AI into a unified gateway for the tool stack

    ⚡ SPARK

    March makes Port AI an MCP gateway: external MCP servers from Atlassian, Notion, GitHub and others plug directly into the side chat, so Port becomes the single AI entry point across the stack. Visual data-source mapping also begins replacing hand-written YAML. The MCP move is the directional one, extending January's ecosystem entry into an aggregation play.

  4. 3mo ago

    Port Product Updates - What we built in February

    February pairs AI expansion with governance hardening: an Azure Anthropic BYO-LLM provider keeps Claude traffic inside a customer's Azure tenant, while a Dynamic RBAC UI and Permissions Simulator move access control out of JSON and let admins test policies before they bite. Solid enterprise-readiness work that supports the AI push rather than redirecting it.

  5. 4mo ago

    Skills, AI Memory, and an official Anthropic MCP connector

    ⚡ SPARK

    January is Port's entry into the agentic ecosystem: Skills encode reusable team expertise for AI and agents, AI Memory adapts to user preferences, and Port ships as an official connector in Anthropic's MCP ecosystem and an approved n8n node. This is the start of the MCP-and-agents arc that March and April build on.

  6. 5mo ago

    Port product updates - What we built in December

    December lays groundwork the AI arc later builds on: an AI integration assistant troubleshoots failed syncs and suggests mapping fixes, Port-hosted custom Ocean integrations remove self-hosting overhead, and Scorecards 2.0 promotes scorecards to first-class catalog entities. Meaningful platform maturation, but mostly deepening existing surfaces rather than opening new ones.