From Search on the Internet to Search within the IDE
Final yr we launched AI-driven semantic search for Meraki API docs on developer.cisco.com—however builders stay of their IDEs, not the browser. With out the appropriate context, AI assistants within the IDE fall again to outdated or generic data when it generates code.
The DevNet Content material Search MCP server is the following step in that evolution: it takes the identical semantic search that powers developer.cisco.com and places it inside your IDE, the place your AI assistant can use it in actual time. MCP (Mannequin Context Protocol) is an ordinary that lets IDE AI assistants name stay instruments—on this case, looking out Cisco API documentation—so that you get correct, present context with out switching home windows. One click on to put in, nothing to configure, and your IDE can now floor its code solutions in official Cisco API specs as a substitute of outdated or generic data.
In the present day the server helps Meraki and Catalyst Middle API documentation; we’re increasing to extra Cisco merchandise and DevNet content material. Need your product’s APIs included? Create a difficulty to request it.
One-Click on Set up, Dwell API Search
The DevNet Content material Search MCP server has one-click set up for VS Code or Cursor—your atmosphere is prepared, nothing to configure. The server doesn’t substitute the browser; it brings the appropriate slice of DevNet into the circulation of coding.
As soon as put in, strive a question out of your IDE chat panel—for instance, “Discover Meraki L3 firewall API endpoints for configuring site visitors guidelines” or “Discover Catalyst Middle APIs for system stock administration”—and see the AI assistant pull stay documentation and use it in its response.
What’s Below the Hood
The MCP server exposes three instruments that your IDE’s AI assistant can name to floor its solutions in DevNet:
- Meraki API Documentation Search — Pure language semantic search over Meraki API documentation. Ask in plain language; the server returns matching endpoints with names, descriptions, API path, technique, operation ID, and a hyperlink to the official doc.
- Meraki API Operation ID Lookup — While you already know the OpenAPI operation ID (e.g.
createNetworkMerakiAuthUser), this instrument returns the exact endpoint particulars and specification. Helpful for validation and for producing code that targets a particular operation. - Catalyst Middle API Documentation Search — The identical semantic seek for Catalyst Middle APIs: enterprise community administration, system onboarding, assurance, compliance, and associated use instances.
Responses are structured in order that the AI assistant can reliably extract api_path, api_method, api_operation_id, and documentation_url and cite them in generated code. The documentation returned is similar supply of fact that powers developer.cisco.com/docs—maintaining outputs tied to the true API floor and making it simpler so that you can confirm what the IDE produces.
Past Search: Reference Workflows and Prompts
The MCP server is most helpful when it’s half of a bigger workflow: fetch the most recent API particulars, then generate or refine code in the identical circulation. To indicate what that appears like, we offer reference pattern workflows and immediate definitions within the GitHub repo—pattern configurations for Meraki and Catalyst Middle that reveal a documentation-first strategy: name the MCP search instruments earlier than producing code, and tailor output to “easy script” vs “enterprise-style” patterns. Instance prompts cowl duties like fetching community IDs, detecting orphan units, or scheduling firmware upgrades.
These are reference samples solely—beginning factors so that you can adapt in your individual atmosphere, not a licensed or production-ready providing.
A word on AI-generated code: Output from any AI assistant—whether or not utilizing this MCP server or not—will be incomplete, incorrect, or insecure. Deal with generated code the way in which you’d a draft from a junior colleague: overview it, take a look at it, and approve it earlier than operating it in any actual atmosphere. We take this significantly, and it shapes the place we’re heading subsequent.
What’s Subsequent
This launch is about one factor: placing the appropriate API context in entrance of the IDE in order that high quality begins from a stable base. The place we go from right here has two tracks.
1. Broader protection. In the present day the server covers Meraki and Catalyst Middle APIs. We’re extending it to extra Cisco product APIs and different DevNet content material—Studying Labs, code samples, sandbox references—so the IDE can use extra of DevNet, not simply API specs.
2. From reference samples to developer expertise. The workflows and prompts we ship immediately are reference samples. The following step is to evolve them into developer expertise: curated, validated configurations with finest practices, safety pointers, and coding requirements in-built. We’re not there but, and we’ll be clear about that development.
Your suggestions will form each tracks—which APIs so as to add first, which use instances to prioritize, and the way we mature the samples. We invite you to strive it, share suggestions, and inform us which product APIs and use instances you’d prefer to see subsequent.
Set up (one-click): VS Code · Cursor
Docs & suggestions: GitHub repo · Request extra APIs