Show HN: AI agents are bad at API integrations – we fixed it(apimatic.io)
apimatic.io
Show HN: AI agents are bad at API integrations – we fixed it
https://www.apimatic.io/product/context-plugins/showcase
4 comments
PayPal benchmark is a smart way to launch this I think...showing real numbers on a recognizable API gives people something concrete to evaluate against. The 65% token reduction is probably the stat that'll matter most to teams watching API costs. One thing I'd be a little worried about is how this handles APIs that version aggressively or have breaking changes between major versions. The context packaging seems like it'd get complicated fast when you're maintaining accuracy across v1 and v2 of the same API simultaneously.
Agents often fall back to outdated docs or SDKs, which breaks real integrations.
Really like the direction here of providing structured, up-to-date API context instead of relying on web search.
The key insight for us was that the context needs to be language-specific and structured around actual SDK methods, not just an API Reference or OpenAPI Spec. Both of those provide context on the shape of the API but leave it to AI agents to figure out exactly how to call API endpoints and then translate that into code - handling auth, errors, pagination, rate limits etc.
That's 2 levels of inference, one of which can be avoided by providing SDKs and context for those SDKs. The SDK wraps all the plumbing for interacting with the API; AI agents just need to figure out which methods to call.
If you get a chance to try one of the examples in the showcase, we'd love to hear your feedback.
That's 2 levels of inference, one of which can be avoided by providing SDKs and context for those SDKs. The SDK wraps all the plumbing for interacting with the API; AI agents just need to figure out which methods to call.
If you get a chance to try one of the examples in the showcase, we'd love to hear your feedback.
This started from a conversation at PayPal DevDay 2025. The PayPal developer experience team were monitoring developers using AI agents to integrate PayPal APIs, and the agents kept reaching for outdated docs and deprecated SDK versions, often falling back to training data instead of reading the current API documentation.
We tested this ourselves with Cursor. Across multiple iterations, 87% of runs fetched outdated reference docs using web search, and 13% of implementations used a deprecated SDK version.
What we built:
We built Context Plugins - given an OpenAPI spec, we generate SDKs and a remote MCP server that exposes language-specific API context (auth flows, method signatures, error handling, pagination, integration patterns from the API team) packaged and optimised for LLM consumption. Instead of guessing from training data or relying on web search, coding agents get the correct SDK methods and integration patterns they need to build API integrations.
We've just launched in Beta with PayPal: https://developer.paypal.com/serversdk/java/getting-started/...
We benchmarked Context Plugins on two codebases, nopCommerce (a mature ASP.NET e-commerce platform) and Microsoft's eShop reference app, and saw 2x faster integration time, 65% lower token usage, and near-zero hallucinations.
Getting started:
- We've published ready-to-use Context Plugins for Spotify, Twilio, Slack, and other APIs; the Show HN link takes you to our product showcase with one-click install for popular AI agents.
- Want to generate Context Plugins from your own spec? Sign up for a free 2-week trial on the APIMatic website, no credit card required.
- Want to dive deeper into the data? Read more here https://www.apimatic.io/product/context-plugins/case-study
We're really looking forward to hearing the community's feedback on this. We're particularly interested in hearing from API platform teams seeing this problem with their users, developers who've fought this problem manually or anyone who's tried building something similar.
What are we missing? What would make this more useful?