Platform · 2025–now · AI Present
Vendor agent APIs
The Codex/Claude loop as a product SKU — durable sessions, compaction, subagents, and a sandbox you do not own.
How Vendor agent APIs mutated
Vendor agent APIs mutated the multi-agent fad into leased infrastructure. OpenAI Agents, Anthropic managed agents, and Bedrock AgentCore absorbed the orchestration tax that open-source swarms could not operate. They stuck because running a reliable tool loop is hard; they fail as strategy when session state and residency sit in someone else's control plane. The durable win is a harness with evals and interrupts; the invoice is platform risk with better streaming.
Cost of the fad
What Vendor agent APIs cost
Teams replaced a weekend of LangGraph glue with a managed session API, then discovered residency, recovery, and kill switches all live on the vendor side of the harness.
Case studies
Patterns
Compare with
Related
Framework · 2023–now
Multi-agent orchestration hype
LangGraph, CrewAI, AutoGPT cosplay — then vendor harnesses shipped the same loop as an API. Open-source swarms faded; leased orchestration stuck.
$ Autonomous agent swarms burned API budgets on coordination loops that a single prompt and a cron job would have handled. Debugging "which agent lied" became the new on-call sport.
Architecture · 2024–now
Context / tool protocols
Standard ways for models to talk to tools and data — MCP after the 2026 stateless rewrite, not one-off plugin snowflakes.
Practice · 2021–now
AI pair programming
Coding agents that edit the repo — now table stakes. Generation outran review capacity; ownership is the bottleneck.
Practice · 2023–now
Agent ops / LLM observability
Tracing, cost caps, and prompt versioning for production LLM features — mostly constrained tool loops, not autonomous agents. Datadog for tokens.
Framework · 2023–now
LLM orchestration frameworks
Chains, tools, memory, and agents as a framework — LangChain-class glue sold as architecture. Fatigue was real; the repos became MCP clients, not tombstones.
$ Teams adopted chain/agent frameworks before they had a single reliable tool call. Abstraction layers multiplied while prompt quality stayed flat.
Practice · 2025–now
Agent skills / SKILL.md
Packaged workflows in a folder — name, description, scripts — loaded on demand instead of stuffing the system prompt.