Open Beats Closed SaaS
Brett Hogan@brett
−4.0%
since Oct 6
The idea
I believe the only real moat left in software is being fully open to machines: API, CLI, MCP, not just a UI. So this basket leans into usage-billed infrastructure that AI agents call directly, plus seat-based tools that have genuinely opened themselves to agents, developer tooling agents work through, and the commerce and identity rails they transact on. It skips companies whose value lives only behind a login screen.
What it owns
- Platforms that charge per API call, per query, per message or per unit of compute earn more as agents replace human users and multiply the volume of machine-driven activity. · CloudflareAgent-driven MCP servers run on Cloudflare's network and get billed by usage, so growth requires machine traffic flowing through Cloudflare at scale., DatadogEvery agent calling APIs generates logs, traces and metrics that Datadog bills by volume, so growth depends on agent activity translating into higher usage-based revenue., MongoDBMongoDB has an official MCP server and Atlas charges by consumption, so growth depends on agents reading and writing through Mongo to drive up usage fees., TwilioTwilio charges per message and per call with no UI to lose, so growth depends on agents multiplying communication volume through its APIs., ElasticElastic sells search and retrieval through APIs on consumption pricing, so growth depends on agents calling Elastic as a default retrieval layer., SnowflakeSnowflake charges per query, so agents running thousands of small queries become paying customers at machine speed if that consumption shows up in revenue., FastlyFastly delivers and accelerates APIs at the edge where agent traffic rides, so growth depends on agent-era API volume translating into edge compute and delivery revenue.46%
- Per-seat software vendors that shipped official machine control protocols can keep their value if agents drive workloads through APIs and integrations rather than through the interface. · AtlassianAtlassian's official MCP server lets agents work in Jira and Confluence without the UI, so growth depends on Atlassian keeping full functionality available through machine interfaces., HubSpotHubSpot moved early on machine control protocols for CRM, so growth depends on agents doing real CRM work through HubSpot's APIs and pulling workloads from closed rivals., AsanaAsana adopted machine control protocols early, so growth depends on agents managing work in it directly and keeping Asana relevant as agents take over coordination., Monday.comIt makes a seat-priced work platform that is opening its data and workflow graph to agents, betting that machine-readable value outlasts UI-gated features.17%
- CLI-native and API-first tools that agents already use to build and ship software will see higher usage and activity as agents become the primary workforce. · GitLabGitLab is API-first end to end and coding agents are already heavy users of DevOps pipelines, so growth depends on agent-written code flowing through GitLab to drive usage and compute revenue., JFrogJFrog's artifact pipeline is CLI-native and built for machines, so growth depends on agents producing more code and artifacts moving through JFrog at scale., PagerDutyPagerDuty runs incident response through APIs, so growth depends on agents operating software around the clock producing more automated incident handling.12%
- When agents act on a user's behalf to transact or log in, these open infrastructure layers collect the traffic and build their businesses on that agent-driven flow. · ShopifyShopify is building a machine-facing checkout layer, so growth depends on agent-initiated purchases becoming real volume on Shopify's rails., OktaEvery agent acting on behalf of a user needs identity and permissions, so growth depends on agent identity becoming a product Okta sells at scale rather than platforms bundling it themselves.11%
- Cash15%