Writing a Coprocessor

This guide walks you through creating your first coprocessor service and connecting it to Hive Router.

At the end, you will have a working setup where Hive Router calls your external service during request processing, and your service can either continue the request or stop it early with a custom response.

Before you start

You need three things:

a running Hive Router
a valid supergraph.graphql
a small HTTP service that accepts and returns JSON

If you do not have a supergraph yet, you can use a test supergraph:

curl -sSL https://federation-demo.theguild.workers.dev/supergraph.graphql > supergraph.graphql

Build your first coprocessor

Create a local service

Hive Router sends a JSON payload with fields like version, stage, control, and optional request data. Your service should return JSON with at least:

Minimal required response

{
  "version": 1,
  "control": "continue"
}

This tells the router to continue processing.

Strict contract

Return OK 200 with valid JSON. Invalid responses are treated as coprocessor failures.

Let's start by creating a small service in any language. It must expose one HTTP endpoint, for example POST /coprocessor.

For local testing, run it on http://127.0.0.1:8081/coprocessor.

Example coprocessor service written in Go

package main

import (
	"encoding/json"
	"log"
	"net/http"
)

type CoprocessorRequest struct {
	Version int                    `json:"version"`
	Stage   string                 `json:"stage"`
	Headers map[string]string      `json:"headers,omitempty"`
	Context map[string]interface{} `json:"context,omitempty"`
}

type CoprocessorResponse struct {
	Version int                    `json:"version"`
	Control interface{}            `json:"control"`
	Headers map[string]string      `json:"headers,omitempty"`
	Body    interface{}            `json:"body,omitempty"`
	Context map[string]interface{} `json:"context,omitempty"`
}

func main() {
	http.HandleFunc("/coprocessor", handleCoprocessor)

	log.Println("Coprocessor running on http://127.0.0.1:8081/coprocessor")
	log.Fatal(http.ListenAndServe("127.0.0.1:8081", nil))
}

func handleCoprocessor(w http.ResponseWriter, r *http.Request) {
	if r.Method != http.MethodPost {
		http.NotFound(w, r)
		return
	}

	var payload CoprocessorRequest

	if err := json.NewDecoder(r.Body).Decode(&payload); err != nil {
		writeJSON(w, CoprocessorResponse{
			Version: 1,
			Control: "continue",
		})
		return
	}

	log.Printf("Received coprocessor stage: %s", payload.Stage)

	writeJSON(w, CoprocessorResponse{
		Version: 1,
		Control: "continue",
	})
}

func writeJSON(w http.ResponseWriter, response CoprocessorResponse) {
	w.Header().Set("content-type", "application/json")
	w.WriteHeader(http.StatusOK)

	if err := json.NewEncoder(w).Encode(response); err != nil {
		log.Printf("failed to write response: %v", err)
	}
}

The example coprocessor handles the router.request stage. It checks whether the incoming request (inbound request to Hive Router instance) includes an Authorization header. If the header is present, it adds "auth.checked": true to the context. If not, it stops the request early with a 401 Unauthorized response.

Configure Hive Router to call your service

Right now, Hive Router is not configured to send anything to the coprocessor. Next step is to add coprocessor to the config file:

router.config.yaml

supergraph:
  source: file
  path: ./supergraph.graphql

coprocessor:
  url: http://127.0.0.1:8081/coprocessor
  protocol: http1
  timeout: 1s
  stages:
    router:
      request:
        include:
          headers: true

This enables one stage (router.request) and sends headers to your service.

To modify context, you don't have to set context: true in include. The same applies to headers or body.

Run Router and validate the call path

Start Hive Router and send any GraphQL request.

If everything is connected, your coprocessor service should receive one JSON payload for each request that reaches router.request.

Add an early auth gate with `control.break`

Now implement a simple policy in your service:

if authorization header exists, return continue and add "auth.checked": true to the context.
if it is missing, return break

Example break response:

Example response

{
  "version": 1,
  "control": { "break": 401 },
  "headers": {
    "content-type": "application/json"
  },
  "body": {
    "errors": [{ "message": "Unauthorized" }]
  }
}

Example coprocessor service written in Go

func handleCoprocessor(w http.ResponseWriter, r *http.Request) {
	if r.Method != http.MethodPost {
		http.NotFound(w, r)
		return
	}

	var payload CoprocessorRequest

	if err := json.NewDecoder(r.Body).Decode(&payload); err != nil {
		writeJSON(w, CoprocessorResponse{
			Version: 1,
			Control: "continue",
		})
		return
	}

	log.Printf("Received coprocessor stage: %s", payload.Stage)

	// Short-circuit with 403 Unauthorized when "Authorization" header is not set.
	if payload.Headers["authorization"] == "" { 																		
		writeJSON(w, CoprocessorResponse{ 																						
			Version: 1, 																																
			Control: map[string]int{ 																										
				"break": http.StatusUnauthorized, 																				
			}, 																																					
			Headers: map[string]string{ 																								
				"content-type": "application/json", 																			
			}, 																																					
			Body: map[string]interface{}{ 																							
				"errors": []map[string]string{ 																						
					{"message": "Unauthorized"}, 																						
				}, 																																				
			}, 																																					
		}) 																																						
		return
	} 																																							

	// Otherwise insert `"auth.checked": true` to context.
	writeJSON(w, CoprocessorResponse{
		Version: 1,
		Control: "continue",
		Context: map[string]interface{}{																							
			"auth.checked": true,																												
		},																																						
	})
}

This stops the pipeline and returns your response to the client immediately.

Operate it safely in production

Because coprocessors are in the request path, treat them like any critical service:

set a strict timeout
monitor request rate, latency, and error rate per stage
use traces and structured logs to debug failures quickly

Edit on GitHub