LoginSign Up
← Back to all sparks
Daily Brief · June 16, 2026

Every sector is opening its product to AI agents at once — MCP and public APIs become table stakes.

agentic-platformsmcppublic-apiagent-infrastructureidentityobservability
O
By Onur Öztürk
LinkedIn ↗
Generated 2h agoDrawn from 17 products

The lead

The day's signal isn't one product — it's a posture every product is adopting at once. Across analytics, devtools, marketing, identity, and ecommerce, the notable releases shared a single move: open the product to AI agents as first-class operators. Fourteen products shipped MCP servers or public APIs; thirty-two touched agent capability in some form. The question has shifted from "does this tool have AI features" to "can an agent drive this tool on a user's behalf, under governance."

Count is the cleanest microcosm: a BI canvas reorganizing itself around a governed, agent-operated core — a public REST API, a hosted MCP server with OAuth and service accounts, and an agent that can read and edit the entire canvas. The same shape recurs at very different scale in GitHub Copilot, which moved Agentic Workflows into public preview, pushing its center of gravity from inline completion toward orchestrated agents that triage, review, and fix across a repo.

What moved

  • Agent access layers landed across the board. Count, Axiom, Customer.io, Kit, Planable, and Document360 all shipped MCP servers or public APIs this cycle — the connective plumbing that lets external agents query and operate each product. Customer.io went further, letting teams author custom agent skills rather than ship a single fixed assistant.
  • Agent infrastructure deepened underneath. Vercel widened its AI Gateway (new Moonshot and DeepSeek models, harness-level APIs in AI SDK 7) and pushed function limits to 30 minutes to host longer-running agentic jobs. Rivet added agentOS, a WASM/V8-isolate VM for agents, atop its actor runtime. The Anthropic TypeScript SDK added Managed Agents deployment support, broadening from a request/response client toward an agent-deployment surface.
  • Identity is being rebuilt for non-human callers. Auth0 clustered its releases on machine-to-machine access and Token Vault, explicitly framing AI agents and partner backends as first-class clients — the governance layer the rest of this shift quietly depends on.
  • Marketing automation is wiring agents into the core. Insider and ConvertKit both shipped spark-level agent and automation work, alongside Customer.io's agent-suggested field mappings for Salesforce and HubSpot onboarding.
  • Observability completed its triad. Axiom made Metrics generally available — unifying logs, traces, and metrics — and exposed all three to agents via MCP, betting the product line on AI engineering.

Sectors today

  • Devtools (8 products): the busiest sector by spark count — Vercel, Unleash, Timely, and Retool all shipped, with the AI Gateway and agent-infra work leading.
  • Development (7): Vercel and Rivet anchored, with Auth0 and HashiCorp adding agent-aware identity and infra.
  • Marketing-automation (5): Insider and ConvertKit led on agent capability; Customer.io threaded agents through setup flows.
  • Marketing (9): Kit and Planable both opened up with public APIs and MCP, moving from point tools toward programmable platforms.
  • Analytics (3): small but dense — Count (3 sparks) and Axiom carried the day's clearest agent-operated narrative.
  • AI-assistants (14): the largest sector by volume; GitHub Copilot, the Anthropic SDK, and LibreChat set the agent-platform tone.
  • Ecommerce (6): Shiprocket led on logistics, with ShipHero and Subbly adding incremental moves.
  • Collaboration (10): high volume, lower spark — Document360's MCP work was the standout.

CRM, communication-messaging, video-conferencing, design, and lms-edtech all saw updates but no spark-level moves — steady maintenance, not direction.

Watch tomorrow

The MCP-and-public-API pattern is the thread to follow. Once a critical mass of these tools is agent-addressable, the next visible move is composition — agents chaining across products (a Customer.io agent reaching a Kit audience, a Copilot workflow querying Axiom). Watch whether Auth0's M2M and Token Vault work starts showing up as the named identity layer in other products' agent stories; identity is the bottleneck this whole shift runs through, and it's the one place where today's many parallel bets have to converge.