Post

🥊 reShapr vs Official GitHub MCP: What happens when you optimize tool calls for AI agents? In our latest benchmark, we put the official @github.com Model Context Protocol (MCP) tools head-to-head with a purpose-built reShapr action to measure pull request velocity on microsoft/vscode. The results?

reshapr.ioreShapr vs Official GitHub MCP 🥊 | reShaprA live, curl-reproducible benchmark showing how reShapr turns a 31-call GitHub MCP pull-request workflow into one predictable agent action, then trims the response further with output filters and…

20/7/26, 2:00 pm0 reposts6 likes

Methiaff

@methiaff.bsky.social

20 July

so the 'official' protocol needed a purpose-built wrapper to actually work efficiently?

Yacine Kheddache

@kheddache.me

20 July

Not the protocol but the official GitHub MCP server, which exposes general-purpose, low-level tools. reShapr takes existing APIs (Rest, gRPC and GraphQL) and exposes them as use-case-level actions tailored to what you actually want the agent to accomplish.

Reshapr

@reshapr.io

20 July

❌ Official @github.com MCP: 31 separate tool calls (1 list + 30 follow-ups) ✅ reShapr: 1 predictable, task-shaped agent action By shaping the API exactly to the agent's needs, we eliminated unnecessary API loops, resulting in a 𝟵𝟳.𝟮% 𝗿𝗲𝗱𝘂𝗰𝘁𝗶𝗼𝗻 𝗶𝗻 𝗿𝗲𝘀𝗽𝗼𝗻𝘀𝗲 𝗯𝘆𝘁𝗲𝘀 and a drastic improvement in latency.

Reshapr

@reshapr.io

20 July

Stop context overload and make your AI agents predictable and reproducible. Check out the full benchmark and methodology here: 👉 reshapr.io/blog/reshapr... #AI #AgenticAI #MCP #GitHub #LLM @aaif.io @linuxfoundation.org 🙌