Skip to content

API DOCUMENTATION CENTER

Provider documentation: models, APIs and integrations.

Access GPT, Claude, Gemini, DeepSeek and Qwen models through one key and compatible endpoints.

Sections11
Procedures54
Formats8
LanguageEN
RECOMMENDED JOURNEY

From an API key to the first request.

Follow these sections to connect your tools and monitor usage.

01CreateCreate a key and choose a model.
02ConnectConfigure OpenAI, Anthropic or Codex.
03MonitorReview logs, pricing and limits.
SERVICE ARCHITECTURE

One gateway, multiple official formats.

ACLIDE routes requests through configured providers while presenting models, costs and limits consistently.

01

API key

Bearer authentication and immediate revocation.

02

Catalog

Models and capabilities in one place.

03

Formats

OpenAI Responses and Anthropic Messages.

04

Routing

Automatic selection of an available route.

05

Logs

Tokens, status and cost for each request.

06

Billing

Credits, prices and plan limits.

Security

Upstream provider keys remain server-side. Never include an API key or sensitive prompt in a support ticket or screenshot.

Check an API key

This check only calls GET /v1/models. The key is never stored and is cleared after the test.

01 / GUIDE

Connect an application

Create a key and send your first request to the available models.

  • Create a key
  • Choose an endpoint
  • Verify a response

Step by step

  1. 01

    Create an API key

    In the dashboard, open API keys, enter a clear name and copy the secret once.

    The secret starts with aclide_sk_. It is not shown again after you close the creation screen.
  2. 02

    Verify the key

    Use the verifier above or call GET /v1/models. A 200 response confirms that the key, account and plan are active.

    cURL
    curl https://aclide.com/v1/models \
      -H \"Authorization: Bearer $ACLIDE_API_KEY\"
  3. 03

    Choose a model

    Copy an identifier marked as available in /v1/models. The catalog depends on your plan.

  4. 04

    Install the server SDK

    For a Node.js website, install the official OpenAI SDK in your backend. Never call ACLIDE directly from browser code.

    Shell
    npm install openai
  5. 05

    Create the ACLIDE client

    Store ACLIDE_API_KEY in your host's secret settings and configure the OpenAI-compatible base URL.

    TypeScript (server)
    import OpenAI from "openai";
    
    const client = new OpenAI({
      apiKey: process.env.ACLIDE_API_KEY,
      baseURL: "https://aclide.com/v1",
    });
  6. 06

    Call Responses from a server route

    Your frontend calls your own /api/chat route; only that server route contacts ACLIDE.

    TypeScript (server)
    const response = await client.responses.create({
      model: "CATALOG_MODEL",
      input: message,
    });
    
    return Response.json({ text: response.output_text });

Tips

  • Keep keys server-side and use the official model catalog.

Checks

  • The configuration is valid
  • No secret is exposed
02 / GUIDE

API keys and security

Manage, rotate and revoke secrets without exposing your credentials.

  • Restrict access
  • Revoke a key
  • Protect secrets

Step by step

  1. 01

    Name each key

    Use one name per project so its activity is easy to identify in logs.

  2. 02

    Revoke immediately

    Open API keys and select Revoke. Every new request using that key is then rejected.

  3. 03

    Keep keys server-side

    Keys must never appear in client code, screenshots or logs.

Tips

  • Keep keys server-side and use the official model catalog.

Checks

  • The configuration is valid
  • No secret is exposed
03 / GUIDE

Model catalog

Explore the available GPT, Claude, Gemini, DeepSeek, Qwen and other model families.

  • Compare models
  • Review capabilities
  • Choose the right plan

Step by step

  1. 01

    Open the catalog

    The Models page lists the identifier, family, capabilities and exact price of every model.

  2. 02

    Filter by capability

    Filter models by reasoning, vision, tools or speed.

  3. 03

    Use the exact identifier

    Copy the displayed public model ID into your requests.

Tips

  • Keep keys server-side and use the official model catalog.

Checks

  • The configuration is valid
  • No secret is exposed
04 / GUIDE

OpenAI format

Connect Responses, Cursor, ChatBox and compatible OpenAI SDKs.

  • Configure the SDK
  • Enable streaming
  • Handle errors

Step by step

  1. 01

    Set the base URL

    Use https://aclide.com/v1 with the OpenAI SDK and your ACLIDE_API_KEY.

  2. 02

    List available models

    Call client.models.list() first and use an identifier allowed for your account.

  3. 03

    Use Responses

    Call client.responses.create with model and input. ACLIDE does not expose Chat Completions.

  4. 04

    Handle status codes

    401 means an invalid or revoked key; 403 an inactive plan; 429 a temporary limit; and 5xx a retryable outage.

Tips

  • Keep keys server-side and use the official model catalog.

Checks

  • The configuration is valid
  • No secret is exposed
05 / GUIDE

Anthropic format

Use the native Messages API with Claude Code, Cline and Anthropic SDKs.

  • Configure Messages
  • Use tools
  • Enable vision

Step by step

  1. 01

    Set the credentials

    Use ANTHROPIC_AUTH_TOKEN for your ACLIDE key and ANTHROPIC_BASE_URL=https://aclide.com.

  2. 02

    Use Messages

    Call POST https://aclide.com/v1/messages with the Authorization header generated by the Anthropic client.

  3. 03

    Choose a Claude model

    Select a Claude model visible in GET /v1/models and allowed by your plan.

  4. 04

    Test before streaming

    Send one request without streaming, then enable stream: true after receiving valid JSON.

Tips

  • Keep keys server-side and use the official model catalog.

Checks

  • The configuration is valid
  • No secret is exposed
06 / GUIDE

Codex CLI

Configure the official Codex package with ACLIDE's Responses-compatible provider.

  • Install Codex
  • Configure config.toml
  • Launch Codex

Step by step

  1. 01

    Install Codex on Windows

    Install the official @openai/codex package with npm, then reopen PowerShell.

    PowerShell
    npm install -g @openai/codex
    codex --version
  2. 02

    Prepare the .codex directory

    Create the user directory and back up any existing configuration files.

    PowerShell
    New-Item -ItemType Directory -Force ~/.codex | Out-Null
    Copy-Item ~/.codex/auth.json ~/.codex/auth.json.backup -ErrorAction SilentlyContinue
    Copy-Item ~/.codex/config.toml ~/.codex/config.toml.backup -ErrorAction SilentlyContinue
  3. 03

    Choose an included GPT model

    In Dashboard > Models, filter included models and copy a GPT identifier. Codex uses the Responses API.

  4. 04

    Create auth.json

    Create auth.json in ~/.codex. Replace YOUR_ACLIDE_KEY with the key created in the dashboard and never commit this file.

    ~/.codex/auth.json
    {
      "OPENAI_API_KEY": "YOUR_ACLIDE_KEY"
    }
  5. 05

    Create config.toml

    Create config.toml in the same directory. This configuration uses ACLIDE's Responses endpoint.

    ~/.codex/config.toml
    model_provider = "aclide"
    model = "gpt-5.6-sol"
    model_reasoning_effort = "high"
    disable_response_storage = true
    preferred_auth_method = "apikey"
    
    [model_providers.aclide]
    name = "ACLIDE"
    base_url = "https://aclide.com/v1"
    wire_api = "responses"
    requires_openai_auth = true
    request_max_retries = 4
    stream_max_retries = 5
  6. 06

    Run a real ACLIDE test

    Run this command in a project directory. The exact response validates the key, model, Responses and streaming.

    PowerShell / Terminal
    codex exec --skip-git-repo-check "Reply exactly: ACLIDE CODEX OK"
  7. 07

    Start Codex

    Launch the interactive interface in any repository. Requests are sent through https://aclide.com/v1/responses.

    PowerShell / Terminal
    codex
  8. 08

    Fix common errors

    Reopen the terminal when the command is missing. For 401, fix auth.json; for 404, choose a model marked Included.

    PowerShell
    npm update -g @openai/codex
    codex --version

Tips

  • Keep keys server-side and use the official model catalog.

Checks

  • The configuration is valid
  • No secret is exposed
07 / GUIDE

OpenCode

Install OpenCode and declare ACLIDE as a custom Responses provider.

  • Install OpenCode
  • Create opencode.json
  • Choose a model

Step by step

  1. 01

    Install OpenCode

    Install the CLI and verify that the command is available.

    PowerShell / Terminal
    npm install -g opencode-ai
    opencode --version
  2. 02

    Load the ACLIDE key

    Keep the key outside the JSON file. OpenCode reads ACLIDE_API_KEY when it starts.

    PowerShell
    $env:ACLIDE_API_KEY = "YOUR_ACLIDE_KEY"
  3. 03

    List allowed models

    Copy the exact identifier of a model available on your plan.

    PowerShell
    $headers = @{ Authorization = "Bearer $env:ACLIDE_API_KEY" }
    (Invoke-RestMethod https://aclide.com/v1/models -Headers $headers).data |
      Where-Object available | Select-Object id
  4. 04

    Create opencode.json

    Add this file to the project. The @ai-sdk/openai package uses ACLIDE's Responses API.

    opencode.json
    {
      "$schema": "https://opencode.ai/config.json",
      "model": "aclide/AVAILABLE_MODEL",
      "provider": {
        "aclide": {
          "npm": "@ai-sdk/openai",
          "name": "ACLIDE",
          "options": {
            "baseURL": "https://aclide.com/v1",
            "apiKey": "{env:ACLIDE_API_KEY}"
          },
          "models": {
            "AVAILABLE_MODEL": { "name": "ACLIDE - primary model" }
          }
        }
      }
    }
  5. 05

    Verify the model

    Start OpenCode, open the model list and select ACLIDE. The configured ID must match the catalog.

    OpenCode
    opencode
    # In OpenCode: /models
  6. 06

    Test and diagnose

    Send a short prompt. For 401 check the key; for 404 correct the model ID; if ACLIDE is missing, validate the JSON.

    PowerShell
    opencode auth list

Tips

  • Keep keys server-side and use the official model catalog.

Checks

  • The configuration is valid
  • No secret is exposed
08 / GUIDE

Claude Code

Connect Claude Code to ACLIDE's compatible Messages endpoint.

  • Choose a client
  • Set the base URL
  • Validate streaming

Step by step

  1. 01

    Install Claude Code on Windows

    Install the official CLI with npm. Claude Code requires Git for Windows or WSL.

    PowerShell
    npm install -g @anthropic-ai/claude-code
    claude --version
    claude doctor
  2. 02

    Prepare the configuration directory

    Create ~/.claude and back up the existing settings file.

    PowerShell
    New-Item -ItemType Directory -Force ~/.claude | Out-Null
    Copy-Item ~/.claude/settings.json ~/.claude/settings.json.backup -ErrorAction SilentlyContinue
  3. 03

    Choose included models

    In Dashboard > Models, filter included Claude models and copy their public identifiers.

  4. 04

    Create settings.json

    Create ~/.claude/settings.json. Claude Code uses ACLIDE's Anthropic Messages endpoint, not the Codex Responses configuration.

    ~/.claude/settings.json
    {
      "env": {
        "ANTHROPIC_API_KEY": "YOUR_ACLIDE_KEY",
        "ANTHROPIC_BASE_URL": "https://aclide.com",
        "ANTHROPIC_MODEL": "claude-sonnet-5",
        "ANTHROPIC_DEFAULT_SONNET_MODEL": "claude-sonnet-5",
        "ANTHROPIC_DEFAULT_OPUS_MODEL": "claude-opus-5",
        "ANTHROPIC_DEFAULT_HAIKU_MODEL": "claude-haiku-4-5-20251001",
        "CLAUDE_CODE_SUBAGENT_MODEL": "claude-haiku-4-5-20251001",
        "CLAUDE_CODE_DISABLE_NONESSENTIAL_TRAFFIC": "1",
        "CLAUDE_CODE_SUBPROCESS_ENV_SCRUB": "1"
      },
      "model": "sonnet",
      "effortLevel": "high"
    }
  5. 05

    Close an existing Anthropic session

    If Claude Code still uses a connected Anthropic subscription, sign out once and restart it.

    Claude Code
    claude auth logout
  6. 06

    Run a real test

    Non-interactive mode must return the requested text exactly. This validates Messages, the key, model and streaming.

    PowerShell / Terminal
    claude -p "Reply exactly: ACLIDE CLAUDE OK"
  7. 07

    Start Claude Code

    Launch Claude Code normally in the project. A 401 means the key is invalid; a model error means it is not included in the plan.

    PowerShell / Terminal
    claude

Tips

  • Keep keys server-side and use the official model catalog.

Checks

  • The configuration is valid
  • No secret is exposed
09 / GUIDE

Logs and usage

Track requests, costs, models and limits from the dashboard.

  • Read a log
  • Understand a cost
  • Diagnose an error

Step by step

  1. 01

    Filter requests

    Filter by key, model, status, period and API format.

  2. 02

    Compare tokens and prices

    Each row shows token counts, the route multiplier and billed cost.

  3. 03

    Analyze a failure

    Use the HTTP status and request ID without sharing the prompt.

Tips

  • Keep keys server-side and use the official model catalog.

Checks

  • The configuration is valid
  • No secret is exposed
10 / GUIDE

Pricing, credits and limits

Understand monthly, quarterly and annual plans, credits and multipliers.

  • Compare plans
  • Review limits
  • Track credits

Step by step

  1. 01

    Choose a billing period

    The monthly, quarterly and annual selector displays the price and discount.

  2. 02

    Understand routing costs

    Each model card displays its exact input, cache and output pricing.

  3. 03

    Monitor the balance

    The dashboard shows remaining credits, renewal and alerts.

Tips

  • Keep keys server-side and use the official model catalog.

Checks

  • The configuration is valid
  • No secret is exposed
11 / GUIDE

API troubleshooting

Quickly resolve authentication, quota, model and network errors.

  • Identify the status code
  • Check the configuration
  • Contact support

Step by step

  1. 01

    Read the HTTP status

    401 invalid key, 402 insufficient balance, 404 unknown endpoint or model, 429 limit, 503 temporary outage.

  2. 02

    Retry a minimal request

    Remove tools and conversation history, then test a basic model.

  3. 03

    Share a safe diagnostic

    Provide the request ID, time and HTTP status, never the key or sensitive content.

Tips

  • Keep keys server-side and use the official model catalog.

Checks

  • The configuration is valid
  • No secret is exposed
QUICK HELP

Essential answers.

Answers to the most common API questions.

Where do I create an API key?

Open API keys in the dashboard and select Create a key.

How do I protect secrets?

Keep them server-side, revoke exposed keys and never place them in Git or browser code.

READY TO START

Connect your first tool.

Create a key, choose a model and send a test request.