Internal developer platforms may need to become agentic engineering platforms as organisations support autonomous AI agents. A new Platform Engineering guide sets out nine practices for that shift, including managed gateways, distinct agent identities and trustworthy platform state.
Watch Desk analysis
What happened
The guide draws on public engineering practices from Block, Uber, LinkedIn, Spotify and major cloud providers. It describes a playbook for supporting agents within internal developer platforms, with structured tool descriptions among the practices it highlights. Read the Platform Engineering guide.
Why it matters
Agents need more than access to tools: operators also need ways to identify them, manage how they connect and give them reliable information about the systems they work in. The guide’s examples make the shift concrete, even if the supplied account does not spell out all nine practices or how each company applies them.
For platform teams, this is a useful framing of the work involved in moving from developer platforms built around people to ones that must also accommodate autonomous software. The word “agentic” is doing plenty of work across the industry; gateways, identities and dependable platform state are rather more tangible.
Our read
The most valuable idea is that agent support is a platform-design problem, not simply a matter of handing an AI another tool. Teams planning deployments should examine how agents are identified, what tools they can reach and whether the platform gives them trustworthy context. The guide offers a starting map, not evidence that one standard architecture will suit every organisation.
What to watch
- Whether platform teams publish concrete examples of these practices in production.
- How agent identities and gateway controls are implemented across different tools.
- Whether trustworthy platform state becomes a defined, testable requirement rather than a reassuring phrase.
Discussion spark: Which should platform teams prioritise first for autonomous agents: distinct identities, controlled tool access or reliable platform state?
Sources and evidence
- Agent-first IDP: how big tech and the clouds are preparing platforms for agents (16 October 2026, 19:53 UTC)
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.