The deal is small enough that the terms would not move anyone’s model of the market, but the direction it points in matters. Enterprises are pushing agents out of pilots and into production. They need a catalog of tools and a runtime that can authorize and audit what those tools do, and the two are merging into one product.
What Arcade Actually Bought
At its core, Smithery solved a discovery and packaging problem. Find a server in one click, run it in another. That is a developer experience win. It also explains why the index grew to tens of thousands of entries covering both MCP servers and, more recently, Anthropic's Agent Skills format.
Arcade sits a layer below that catalog. The company sells what it calls a secure action layer, which handles delegated authorization so an agent acts as a specific authenticated user with only that user's permissions. Every action it takes leaves an audit record. Arcade says it authored the MCP authorization specification that major clients and servers now reference.
Arcade has the balance sheet to back the ambition. It announced a $60 million Series A in June, led by SYN Ventures with strategic investment from Morgan Stanley and Wipro, taking total funding to $72 million. Combine the registry with the runtime and the pitch to a CIO becomes simple. One vendor finds the tool, runs it, and proves afterward which user it acted for. This is the first time an MCP registry with real developer mindshare has been bought by the layer that governs execution.
Play Puzzles & Games on Forbes
The Quality Gap Arcade Is Pointing At
The rationale Arcade gives rests on its own measurement. Arcade says its ToolBench benchmark indexed more than 43,400 MCP servers and analyzed 219,069 tools. The company reports that 0.5% of them earned an A grade, on a scale running from A+ down to F, based on how a server's tools behave when a model actually calls them.
Read that figure with caution, since Arcade writes the rubric and sells the remedy. The underlying observation is easier to confirm. Anyone who has wired half a dozen community MCP servers into a production agent has watched malformed schemas, missing tool descriptions and overlapping tool names push the model into the wrong call. The rate of publishing has outrun the rate of engineering.
Registries, Runtimes And The Official Catalog
Smithery was never the only place to look. The maintainers launched the official MCP Registry in preview in September 2025 and described it as a community-owned source of truth. It was built deliberately thin, meant for downstream catalogs to mirror rather than to serve as anyone's front end.
Docker moved in a different direction with its MCP Catalog and Toolkit, containerizing servers and wrapping them in secret management, policy enforcement and audit logging. Anthropic then donated the protocol itself to the Agentic AI Foundation under the Linux Foundation, with Block, OpenAI, Google, Microsoft and AWS supporting the move.
The key difference among the three is where trust gets enforced. The official registry verifies namespaces and publishes metadata, and it stops there by design. Docker isolates the process and constrains what it can reach on the host. Arcade wants the request path at execution time, which is the only position from which a security team can answer which agent did what, against which system, on whose behalf.
What The Deal Does Not Fix
Undisclosed terms make this hard to size. A registry with heavy traffic and thin revenue can be worth a great deal strategically and very little financially. Nothing in the announcement distinguishes a talent acquisition from a substantial purchase.
Neutrality is the sharper problem, because developers adopted Smithery partly on the understanding that it was not owned by a vendor selling the layer underneath. Publishers now list their servers in a catalog controlled by a company whose commercial interest is to run those servers on its own runtime. Arcade has not yet addressed that conflict publicly. How ranking, verification and grading stay insulated from its own runtime business is an open question.
The protocol is also moving under everyone. The maintainers used the 2026-07-28 release to remove the initialize handshake and session identifiers for remote servers. That is a breaking change, and both registries and runtimes have to absorb the churn. A grade assigned before the change may not describe how a server behaves after it.
What Enterprise Buyers Should Ask Now
The first question for a platform team is portability. If you standardize on Arcade for both discovery and execution, the tool definitions, the authorization policies and the audit records end up inside one vendor's control plane. Ask what format those artifacts export in, and whether another runtime can consume them without a rewrite.
The second question is whether the grading stays independent. Ask whether ToolBench scores servers that compete with Arcade's own catalog on identical criteria. Ask whether the methodology is published in enough detail to reproduce, and who arbitrates a disputed grade.
The third question is pricing under load. Tool call volume is not stable, and Arcade says its own call volume grew 25 times in six months. Find out what the bill does when one workflow triples its tool calls after a model upgrade, because that is the scenario that turns a modest platform fee into a line item the CFO wants explained.
MCP tooling is no longer something a platform team assembles from free parts. It is becoming a procured layer, with vendors, benchmarks and contracts attached, and Arcade has moved earlier than most to occupy the whole of it. Developers who valued Smithery for its speed should keep getting that speed. Security teams who have been blocking agents from production now have a shorter list of vendors to evaluate before they stop saying no. %!s()
