Skip to content

Open-source agent server generator

Open-source CLI that turns an existing codebase into a working agent-protocol server, generated from the code you already wrote.

Run the CLI in your repo and it reads type signatures, docstrings and help text, proposes tools described for a model to read, and generates the server with a test harness. For developers who want existing code agent-callable.

No account. No card. No sales call.

The problem

Reading a protocol spec to expose eight functions you already have is pure tax.

Most developers abandon it, and the ones who do it produce servers with badly described tools, no input validation and confusing errors, so the agent using them behaves unpredictably and the developer blames the model.

Who it is for

  • Individual developers

    Developers with an existing CLI, library, REST service or internal tool they want agents to be able to call.

  • Small teams

    Teams who want their tools agent-callable without hand-writing a server or learning a protocol spec.

Generate a production config you can copy

Pick authentication, region, client and limits. Copy a config and client snippet that run, with the latency they imply. Runs in your browser. No account, no card, no call.

Your service

Generates locally. Copy the output and it runs.

p50 latency
25 ms
p99 latency
80 ms
Worst case
30.3 s
all retries exhausted
my-service.config.yaml yaml
# my-service — generated by ACPtify
# Copy to my-service.config.yaml. Every value below is live, not a placeholder.
apiVersion: v1
kind: Service
metadata:
  name: my-service
  region: us-east
spec:
  endpoint: https://my-service.acptify.com/v1
  auth:
    mode: apikey
    credential: ${MY_SERVICE_TOKEN}
  limits:
    requestsPerMinute: 600
    timeoutMs: 10000
    maxRetries: 2
    backoff: exponential
  observability:
    structuredLogs: true
    tracing: w3c
    replayRetentionDays: 7
Try it now bash
curl -sS https://my-service.acptify.com/v1/run \
  -H "X-API-Key: $TOKEN" \
  -H "Content-Type: application/json" \
  --max-time 10.0 \
  -d '{"input": {"task": "hello"}}'
Client node
import { Client } from "@my-service/sdk";

const client = new Client({
  endpoint: "https://my-service.acptify.com/v1",
  auth: { mode: "apikey", token: process.env.TOKEN! },
  timeoutMs: 10000,
  maxRetries: 2,
});

const result = await client.run({ task: "hello" });
console.log(result.data);
Things this configuration will do to you
  • API keys do not expire on their own. Set a rotation reminder now.

Take this with you

No email required. It is your result.

ACPtify — config
· my-service · us-east · apikey
· p50 25ms, p99 80ms
· 1,500,000 requests/month at 600 rpm burst

Run it yourself: https://acptify.com

An agent-protocol server, without reading the spec

A code-first generator that reads your repo and writes the server, with tool descriptions written for a model to read.

  • Reads your code

    Run it in a repo and it reads type signatures, docstrings and CLI help text to work out what your tools are.

  • Descriptions written for a model

    It proposes a tool surface with descriptions written for a model to read, not copied from a comment.

  • A generated server

    It generates a working agent-protocol server from that tool surface.

  • Tested with a real model

    The included test harness drives your tools with a real model and reports where a description was ambiguous.

  • Your code, in your repo

    The generated code is committed to your repo and has no runtime dependency on ACPtify.

  • Hosting if you want it

    The paid layer is hosting, versioned deploys and usage analytics, for people who would rather not run the server themselves.

Pricing

The CLI is free and open source. Pay only if you want the servers hosted for you.

CLI

Generate and test servers from your own code.

Freeforever, MIT licensed
  • Code-first generator
  • Test harness
  • No runtime dependency
Generate a config
Recommended

Cloud

Hosting for developers who do not want to run it.

$19per server, per month
  • Usage analytics
  • Versioned deploys
  • Public tool catalogue page
Get in touch

Teams

For teams shipping servers together.

$149per month
  • Private registries
  • CI integration
Talk to us

Prices in USD.

How it works

ACPtify, from the first step to the result.

  1. 01

    Run the CLI in your repo

    It reads your type signatures, docstrings and help text.

  2. 02

    Review the proposed tools

    Check the tool surface and the descriptions it has written for a model.

  3. 03

    Generate and test

    Generate the server, then let the test harness drive it with a real model and flag ambiguous descriptions.

  4. 04

    Commit it, or host it

    Keep the generated code in your repo, or deploy it to ACPtify cloud hosting.

Questions people actually ask

Is the CLI really free?

Yes. The CLI is free forever and MIT licensed.

What does it need from my code?

It reads type signatures, docstrings and CLI help text, and uses them to propose the tools and their descriptions.

Do I depend on ACPtify at runtime?

No. The generated code is yours, committed to your repo, with no runtime dependency on ACPtify.

How do I know the tool descriptions work?

The included test harness drives your tools with a real model and reports where a description was ambiguous, so you can fix it before an agent misuses the tool.

What does hosting cost?

Cloud hosting is 19 dollars a month per server, with usage analytics, versioned deploys and a public tool catalogue page.

Is there a plan for teams?

Yes. 149 dollars a month adds private registries and CI integration.

Why should I trust the config generator?

The config is built from the options you pick, and it lists anything in it that will misbehave under load before you copy it.

Generate a production config you can copy

Pick authentication, region, client and limits. Copy a config and client snippet that run, with the latency they imply. Runs in your browser. No account, no card, no call.

Open the free tool

It runs in your browser. ACPtify never sees your inputs.