Skip to content
BranchesTags

Latest commit

 

History

14 Commits

Folders and files

NameName
Last commit message
Last commit date
.gitignore
.gitignore
 
 
LICENSE
LICENSE
 
 
README.md
README.md
 
 
index.ts
index.ts
 
 
openclaw.plugin.json
openclaw.plugin.json
 
 
package-lock.json
package-lock.json
 
 
package.json
package.json
 
 
tsconfig.json
tsconfig.json
 
 

Repository files navigation

AbacusAI Auth (OpenClaw Plugin)

Third-party provider plugin that integrates AbacusAI models into OpenClaw via direct connection to AbacusAI's RouteLLM endpoint. The plugin configures core compatibility options (requiresAdditionalPropertiesFalse, supportsStrictMode) so that the OpenClaw Agent can use AbacusAI-hosted models (Claude, Gemini, GPT, DeepSeek, Qwen, Grok, Kimi, Llama, and more) with full multi-tool calling support.

Field Value
Package openclaw-abacusai-auth
Entry ./index.ts
Provider ID abacusai
Aliases abacus, abacus-ai, abacusai-code-mode
API style openai-completions (direct connection)
Upstream https://routellm.abacus.ai/v1

Installation

Option 1: Install from npm (recommended)

openclaw plugins install openclaw-abacusai-auth

Option 2: Manual installation

  1. Clone this repository:
git clone https://github.com/tonyhu2006/openclaw-abacusai-auth.git ~/.openclaw/extensions/abacusai-auth
  1. Install dependencies:
cd ~/.openclaw/extensions/abacusai-auth
npm install
  1. Enable the plugin:
openclaw plugins enable abacusai-auth

Quick Start

1. Authenticate

openclaw models auth login --provider abacusai --set-default

The interactive login flow will:

  1. Attempt to auto-detect credentials from a local AbacusAI Code Mode installation.
  2. Fall back to the ABACUSAI_API_KEY environment variable.
  3. Prompt for manual entry if neither is found.
  4. Validate the API key against https://api.abacus.ai/api/v0/describeUser.
  5. Let you select which models to register (defaults to all supported models).
  6. Write the provider config to openclaw.json with compat options.

2. Restart the Gateway

openclaw gateway run

3. Use AbacusAI models

openclaw send "Hello" --model abacusai/gemini-3-flash-preview

Architecture

┌──────────────────────────────────────────────────────────────────┐
│  OpenClaw Agent (Pi Agent)                                       │
│  Sends standard OpenAI-compatible requests                       │
│  (POST /v1/chat/completions with tools[])                        │
└──────────────┬───────────────────────────────────────────────────┘
               │
               ▼
┌──────────────────────────────────────────────────────────────────┐
│  Local Proxy (http://127.0.0.1:<dynamic-port>)                   │
│                                                                  │
│  Schema Normalization:                                           │
│  1. Removes `strict` field from tool definitions                 │
│  2. Removes unsupported keywords (patternProperties, $ref, etc.) │
│  3. Adds `additionalProperties: false` to object schemas         │
│  4. Normalizes SSE responses (adds missing id/object fields)     │
└──────────────┬───────────────────────────────────────────────────┘
               │ https://routellm.abacus.ai/v1
               ▼
┌──────────────────────────────────────────────────────────────────┐
│  AbacusAI RouteLLM Endpoint                                      │
│  OpenAI-compatible API with function calling                     │
│  Routes to Claude, Gemini, GPT, DeepSeek, Llama, etc.           │
└──────────────────────────────────────────────────────────────────┘

Security: The local proxy uses a dynamic port assigned by the OS at startup, avoiding port conflicts and improving security.

Why a local proxy? AbacusAI's RouteLLM is mostly OpenAI-compatible but has strict schema requirements that OpenClaw's default tool schemas don't meet:

  1. Rejects the strict field in tool schemas
  2. Rejects patternProperties and other advanced JSON Schema keywords
  3. Requires additionalProperties: false in object schemas

The local proxy handles all schema normalization internally, making the plugin fully self-contained and compatible with any OpenClaw version

Supported Models

The following models are registered by default (verified February 2026):

Model ID Family
gemini-3-flash-preview Google Gemini
gemini-3-pro-preview Google Gemini
gemini-2.5-flash Google Gemini
gemini-2.5-pro Google Gemini
gpt-5.2 OpenAI GPT
gpt-5.1 OpenAI GPT
gpt-5-mini OpenAI GPT
claude-sonnet-4-5-20250929 Anthropic Claude
claude-opus-4-6 Anthropic Claude
claude-haiku-4-5-20251001 Anthropic Claude
deepseek-ai/DeepSeek-V3.2 DeepSeek
deepseek-ai/DeepSeek-R1 DeepSeek
kimi-k2.5 Moonshot Kimi
qwen3-max Alibaba Qwen
grok-4-1-fast-non-reasoning xAI Grok
route-llm AbacusAI Auto-Router

All models are configured with:

  • Context window: 200,000 tokens
  • Max output tokens: 8,192 tokens
  • Input modalities: text, image
  • API: openai-completions

You can customize the model list during the interactive login flow.

Credential Resolution

The plugin resolves API keys using a multi-tier fallback strategy, checked in order:

During Login (openclaw models auth login)

  1. Local AbacusAI Code Mode installation — scans platform-specific paths:
    • Windows: %APPDATA%\AbacusAI\User\globalStorage\credentials.json, %APPDATA%\AbacusAI Code Mode\User\globalStorage\credentials.json, %USERPROFILE%\.abacusai\credentials.json, %USERPROFILE%\.abacusai\config.json
    • macOS: ~/Library/Application Support/AbacusAI/..., ~/.abacusai/...
    • Linux: ~/.config/AbacusAI/..., ~/.abacusai/...
    • Accepts fields: apiKey, api_key, token, accessToken, access_token
  2. Environment variable — ABACUSAI_API_KEY
  3. Manual entry — interactive prompt

Core Compatibility Options

This plugin configures two core ModelCompatConfig options that are essential for AbacusAI RouteLLM compatibility:

requiresAdditionalPropertiesFalse

AbacusAI RouteLLM requires additionalProperties: false in tool parameter schemas. When this option is set to true, the normalizeToolParameters function in pi-tools.schema.ts sets additionalProperties: false instead of the default true.

supportsStrictMode

AbacusAI RouteLLM rejects the strict field in tool definitions. When this option is set to false, the pi-ai library omits the strict field from tool definitions.

Provider Configuration

The plugin configures the AbacusAI provider to use the local proxy:

{
  "baseUrl": "http://127.0.0.1:<dynamic-port>",
  "api": "openai-completions",
  "auth": "token",
  "compat": {
    "requiresAdditionalPropertiesFalse": true,
    "supportsStrictMode": false
  }
}

The local proxy automatically starts when the plugin loads with a dynamic port assigned by the OS, and handles all schema normalization before forwarding requests to https://routellm.abacus.ai/v1.

Configuration Reference

After login, the plugin writes the following to ~/.openclaw/openclaw.json:

{
  "models": {
    "providers": {
      "abacusai": {
        "baseUrl": "http://127.0.0.1:<dynamic-port>",
        "api": "openai-completions",
        "auth": "token",
        "compat": {
          "requiresAdditionalPropertiesFalse": true,
          "supportsStrictMode": false,
        },
        "models": [
          {
            "id": "gemini-3-flash-preview",
            "name": "gemini-3-flash-preview",
            "api": "openai-completions",
            "reasoning": false,
            "input": ["text", "image"],
            "cost": { "input": 0, "output": 0, "cacheRead": 0, "cacheWrite": 0 },
            "contextWindow": 200000,
            "maxTokens": 8192,
          },
          // ... other models
        ],
      },
    },
  },
}

Credentials are stored separately in ~/.openclaw/agents/<agent>/agent/auth-profiles.json:

{
  "profiles": {
    "abacusai:<email-or-default>": {
      "type": "token",
      "provider": "abacusai",
      "token": "<api-key>",
    },
  },
}

Environment Variables

Variable Description
ABACUSAI_API_KEY API key fallback (used if no saved profile is found)
OPENCLAW_STATE_DIR Override the OpenClaw state directory (default: ~/.openclaw)

Troubleshooting

Tool calling not working

If tool calls fail with schema-related errors, ensure the provider has the correct compat options configured:

"compat": {
  "requiresAdditionalPropertiesFalse": true,
  "supportsStrictMode": false
}

API key validation failed

Your API key may have been revoked or expired. Generate a new one at https://abacus.ai/app/profile/apikey and re-authenticate:

openclaw models auth login --provider abacusai --set-default

Plugin not found

If AbacusAI models are not available, ensure the plugin is installed:

openclaw plugins install openclaw-abacusai-auth

Webchat no response (Zombie Proxy)

If you send messages in Webchat and receive no response, the Gateway may be attempting to communicate with a stale proxy port. This occurs when OpenClaw's internal openclaw.json registers 18862, but a zombie plugin process from a previous run is holding 18862, forcing the new plugin to dynamically increment its port to 18863 or higher. The Gateway sends requests to 18862 (the zombie proxy, which doesn't know how to route current requests), resulting in hanging chats.

Solution (v1.2.8+): The plugin now implements a self-healing /__kill HTTP endpoint. If the plugin detects EADDRINUSE during startup on 18862, it aggressively kills the legacy zombie proxy holding the port, waits 1 second, and successfully binds to 18862. You should no longer experience Webchat unresponsiveness due to port drifting. If it ever hangs, simply restart your Gateway using standard Ctrl+C or openclaw gateway stop to ensure the hook securely cleans up the process.

Getting an API Key

  1. Sign in at https://abacus.ai
  2. Navigate to Profile → API Keys (https://abacus.ai/app/profile/apikey)
  3. Click Generate new API Key
  4. Copy the key (starts with s2_...)

File Structure

openclaw-abacusai-auth/
├── index.ts              # Plugin source (auth, credential detection, model definitions)
├── package.json          # Package metadata
├── openclaw.plugin.json  # Plugin manifest
└── README.md             # This file

Key Constants

Constant Value Description
ROUTELLM_BASE https://routellm.abacus.ai/v1 Upstream RouteLLM endpoint
ABACUS_API https://api.abacus.ai/api/v0 AbacusAI REST API (for key validation)
DEFAULT_CONTEXT_WINDOW 200,000 Default context window for all models
DEFAULT_MAX_TOKENS 8,192 Default max output tokens

License

MIT

Author

tonyhu2006

Contributing

Issues and PRs welcome at https://github.com/tonyhu2006/openclaw-abacusai-auth

About

OpenClaw AbacusAI provider plugin - Third-party plugin for AbacusAI RouteLLM integration

Resources

Readme
MIT license
Activity

Stars

3 stars

Watchers

1 watching

Forks

0 forks
Report repository

Releases

Packages

Contributors

Languages