ENDPOINT DOCTOR / LOCAL-FIRST CLI

Test Your AI API
Before You Ship.

Validate compatibility, capability and reliability of OpenAI-compatible AI APIs.

START IN YOUR TERMINAL
pip install apiwells

Runs locally. Your API keys stay off this website.

DIAGNOSTIC CAPABILITIES3 AREAS
YOUR MACHINEEndpoint Doctor$ apiwells doctor
LOCAL
OpenAI-compatible API
01ConnectivityDNS · TLS · HTTP
02CompatibilityAuthentication · /models · Chat
03CapabilityStreaming · Tools · Structured output
Diagnostic evidence, in your terminal.

01 / THE PROBLEM

AI APIs Fail in
Unexpected Ways.

An endpoint URL is a starting point. Understanding its behavior comes next.

01

Compatibility

Does the endpoint actually follow expected API behavior?

02

Capability

Does it support streaming, tools and structured outputs?

03

Reliability

Can developers understand failures before production?

02 / DIAGNOSTIC COVERAGE

From connection
to capability.

Check connectivity in Basic mode.
Opt into Deep for chat, streaming, tools and structured output.

  • 01

    DNS

    Hostname resolution

  • 02

    TLS

    Secure connections

  • 03

    HTTP

    Response behavior

  • 04

    Authentication

    Access to the endpoint

  • 05

    /models

    Model availability

  • 06

    Chat Completions

    Expected API behavior

  • 07

    Streaming / SSE

    Streaming responses

  • 08

    Usage

    Token usage reporting

  • 09

    Tool Calling

    Tool call capability

  • 10

    Structured Output

    Structured responses

03 / HOW IT WORKS

Your terminal. Three steps.

A short path from an endpoint to diagnostic evidence.

  1. 01

    Install

    Start with the Endpoint Doctor CLI.

    INSTALL
    pip install apiwells
  2. 02

    Run

    Run diagnostics from your machine.

    RUN
    apiwells doctor --base-url https://YOUR-API-HOST/v1
  3. 03

    Understand

    Inspect the evidence before you ship.

    Receive diagnostic report

Replace https://YOUR-API-HOST/v1 with your exact API base URL and configure APIWELLS_API_KEY locally. Capability checks require --deep --model YOUR-MODEL-ID and may incur API charges. Read the quick start

04 / THE OUTPUT

Less guessing.
More evidence.

A real diagnostic run, from connection checks to validated tool calls and structured output.

CLI 0.2.1qwen3.8-flash
  • See the outcome of each check.
  • Inspect timing and capability evidence.
  • Use the evidence to guide your next step.
RECORDED DIAGNOSTIC REPORT
endpoint-doctorRECORDED RUN

Endpoint Doctor Report

CLI 0.2.1 / qwen3.8-flash

PASS

10 checks passed in this recorded run.

URLHTTPS base URL validated
PASS
DNS31 ms / 4 addresses resolved
PASS
TLS1,234 ms / TLSv1.3
PASS
HTTP219 ms / HTTP 200
PASS
Authentication219 ms / credentials accepted
PASSSUPPORTED
/models219 ms / 261 models returned
PASSSUPPORTED
Chat Completions1,156 ms total / usage reported
PASSSUPPORTED
Streaming / SSE2,047 ms total / 12 chunks
PASSSUPPORTED
Tool Calling3,281 ms total / 2-request round trip
PASSSUPPORTED
Structured Output3,578 ms total / schema validated
PASSSUPPORTED
Streaming TTFT1,328 msTime to first token
Chat token usage88 tokens64 input / 24 output
View request evidence
TLS cipher
TLS_AES_256_GCM_SHA384
Target model found
true
Streaming token usage
64 input / 27 output / 91 total
SSE terminal event
true
Tool round trip completed
true
JSON Schema validation
true

Report complete.

Real output supplied by the project owner, condensed for display. One recorded run; timings vary by endpoint and environment. Host and IP details omitted.

05 / BUILT FOR DEVELOPERS

For the people
shipping with AI APIs.

A focused tool for the work between choosing an API and putting it into production.

06 / FUTURE DIRECTION

Start with understanding.

Endpoint Doctor is the starting point.
Evaluation and optimization are long-term directions, not available products.

  1. Endpoint DoctorCURRENT
  2. AI Workload EvaluationFUTURE
  3. Optimization ToolsFUTURE

BEFORE THE NEXT DEPLOY

Know your endpoint.
Then ship.

Start with a diagnostic report from your own machine.