Before AI can work,
it needs to know the place.

At Checkmarx, developers had five ways to create a development environment. Five sounds like choice. In practice, it meant tickets, inconsistent setups and cloud resources left running after everyone went home. One unreliable approach could interrupt the DevOps team’s work. The bottleneck was the machinery around writing software.

Five ways to do the same thing

Checkmarx used Port to offer self-service environment creation and destruction, with a time-to-live that could trigger automatic cleanup. It also replaced collections of AWS resources with bundled k3d environments. Its reported hourly cost fell from $30 to as little as $0.70. That distinction matters: the savings came from redesigning the environment and managing its lifecycle together. A portal cannot make an expensive architecture cheap merely by giving it a button.

That example explains Port better than the phrase “agentic SDLC platform.” Software development lifecycle, abbreviated SDLC, covers the work from an idea to running software. Port began with an internal developer portal: somewhere engineers could understand their systems and do useful things without asking another team to do every step.

The map becomes the machine

Founded in 2022 by Zohar Einy and Yonatan Boguslavsky, Port grew from a problem they had already encountered. The pair had worked on a portal serving more than 2,000 developers in Israel’s Unit 8200. Their commercial company now serves hundreds of organizations globally. Public customers span software, telecoms and payments, including GitHub, British Telecom and dLocal.

The technical idea is a configurable model of an engineering organization. Port calls its unified knowledge layer the Context Lake. Repositories, services, incidents and cloud resources become connected records rather than isolated entries in separate tools. A service can have an owner, dependencies and a deployment history. The relationships make the inventory useful when someone needs an answer.

On top sit workflows. A developer can request an environment; an automation can respond to an event; an agent can participate in a sequence that includes checks and approvals. Port’s workflow orchestrator supports multi-step operations and human involvement through Slack and Teams. Its agent management offering catalogs agents and reusable skills. Governance defines access and gates actions against policy.

The controls are part of the product’s appeal to larger engineering organizations. A service scorecard can express a security or reliability threshold. An access policy can restrict sensitive context. A risky action can be routed to an appropriate person for approval. Those choices still belong to the organization: the software supplies a place to express and enforce them. If ownership is ambiguous or the rules remain unwritten, installing a portal leaves the hard decisions waiting. That is why the data model and the operating habits deserve as much attention as the agent demonstration.

This puts Port beside coding assistants rather than making it another editor. Its wager is that generating code leaves a substantial coordination problem unsolved. An agent still needs to know which service matters, which standards apply and whether it has permission to act. Port sells the context and operating structure around that work.

GitHub changed the order

GitHub already had a homegrown portal and established compliance scorecards. According to its Port customer story, adding a resource type could take three to four months. “It’s not our main focus,” engineering vice president Meirav Feiler said of maintaining that tooling.

Port’s flexible blueprints appealed because GitHub wanted to model its own environment. But the implementation lesson is more interesting than the purchase. GitHub initially planned to finish inventory, add scorecards, then bring in users. It changed course, starting with service experiences while the broader model was still developing. The practical lesson: put something useful in people’s hands before trying to describe the whole company.

The beaver and the key ring

Port has been testing its own argument. Its internal Slack interface, Platform Beaver, started in May 2026 with explicit commands. New commands required code changes; people needed to remember syntax. Natural language made the bot easier to use. Configurable routing then replaced hardcoded decisions about which agent should answer which channel.

The more revealing discovery concerned identity. A shared bot account obscured who had requested an action and prevented permissions from following the individual. Per-user OAuth supplied that missing link. Port reports that narrow work, such as bug triage and incident support, suited agents better than open-ended questions requiring unwritten team knowledge. That is a useful constraint for anyone copying the approach: choose a bounded task, a named owner and a real permission model.

“Developers are still accountable.”Alberto Almeida, CTO, dLocal

At payments company dLocal, the dCoder agent pulls engineering context, plans work, writes code, runs checks and opens a pull request for developer review. A Port-published story reports coverage of 45% of engineering tickets. The company also cleaned and enriched its data and documented architecture decisions. Treat that figure as a customer-reported outcome, dependent on the surrounding work, rather than a forecast for every buyer.

Who pays for the map?

Port makes money through software subscriptions. Its public pricing offers a free tier for up to 15 seats. Basic starts at $30 per seat monthly, sold as a 50-seat package; Standard starts at $40, with Enterprise quoted separately. Capacity limits apply. Platform teams must also budget effort for integrations, data ownership and workflow design.

The alternatives include Backstage, Cortex, OpsLevel and maintaining a homegrown portal. Port’s position rests on adaptable modeling, self-service and an increasingly explicit layer for agents. Its open-source Ocean framework supplies integration plumbing. A January 2026 n8n partnership connects Port’s context and controls to external automation.

Investors have funded that expansion: a $100 million Series C in December 2025 valued Port at $800 million. The ambition is large. The useful starting point remains small: one service with an owner, one action with permission, one less ticket waiting for somebody to notice it.