Openclaw JSON Editing

OpenClaw JSON Editing

Expert guidance for editing JSON in the OpenClaw ecosystem. OpenClaw uses JSON5 for configuration (allows comments, trailing commas), has sophisticated config merging, and validates with Zod schemas.

Quick Reference

OpenClaw JSON5 Config

OpenClaw config files use JSON5 (not strict JSON):

{
  // Single-line comments are allowed
  "gateway": {
    "mode": "http",  // Trailing commas are allowed
  },
  /* Multi-line comments
     are also supported */
  "agents": {
    "main": {
      "model": "anthropic/claude-opus-4-6",
    },
  },
}

Key Differences from JSON

• Comments: Single-line (//) and multi-line (/* */)
• Trailing commas: Allowed in arrays and objects
• Unquoted keys: { key: "value" } is valid
• Single quotes: 'string' is valid

Config File Locations

Safe JSON Operations

Reading Config Files

OpenClaw uses JSON5.parse() for configs and safe wrappers:

// OpenClaw's safeParseJson pattern
function safeParseJson<T>(raw: string): T | null {
  try {
    return JSON.parse(raw) as T;
  } catch {
    return null;
  }
}

// For OpenClaw configs, use JSON5
import JSON5 from "json5";

function loadConfigFile(path: string): unknown {
  try {
    const raw = fs.readFileSync(path, "utf8");
    return JSON5.parse(raw);  // Allows comments, trailing commas
  } catch {
    return undefined;
  }
}

Writing Config Files

OpenClaw writes with specific formatting and permissions:

function saveJsonFile(pathname: string, data: unknown) {
  const dir = path.dirname(pathname);
  if (!fs.existsSync(dir)) {
    fs.mkdirSync(dir, { recursive: true, mode: 0o700 });
  }
  // 2-space indentation, trailing newline
  fs.writeFileSync(pathname, `${JSON.stringify(data, null, 2)}\n`, "utf8");
  fs.chmodSync(pathname, 0o600);  // User read/write only
}

Type Guards

Always validate before assuming structure:

// OpenClaw's isPlainObject (strictest)
function isPlainObject(value: unknown): value is Record<string, unknown> {
  return (
    typeof value === "object" &&
    value !== null &&
    !Array.isArray(value) &&
    Object.prototype.toString.call(value) === "[object Object]"
  );
}

// Less strict version
function isRecord(value: unknown): value is Record<string, unknown> {
  return typeof value === "object" && value !== null && !Array.isArray(value);
}

Config Merging & Patching

Merge Patch (RFC 7386)

OpenClaw uses merge patching for config updates:

// Apply a merge patch to base config
function applyMergePatch(base: unknown, patch: unknown): unknown {
  if (!isPlainObject(patch)) {
    return patch;
  }

  const result: Record<string, unknown> = isPlainObject(base) ? { ...base } : {};

  for (const [key, value] of Object.entries(patch)) {
    if (value === null) {
      delete result[key];  // null = delete key
      continue;
    }
    if (isPlainObject(value)) {
      const baseValue = result[key];
      result[key] = applyMergePatch(
        isPlainObject(baseValue) ? baseValue : {},
        value
      );
      continue;
    }
    result[key] = value;
  }

  return result;
}

Usage Examples

// Add/update nested field
const patch = {
  agents: {
    main: {
      model: "anthropic/claude-opus-4-6"
    }
  }
};

// Delete a field (set to null)
const deletePatch = {
  agents: {
    main: {
      temperature: null  // Removes temperature
    }
  }
};

// Replace entire section
const replacePatch = {
  channels: {
    telegram: null,  // Delete old
    discord: { token: "new-token" }  // Add new
  }
};

Environment Variable Substitution

OpenClaw configs support ${VAR} and ${VAR:-default} syntax:

{
  "auth": {
    "profiles": {
      "openai": {
        "apiKey": "${OPENAI_API_KEY}"  // Substituted at load time
      },
      "anthropic": {
        "apiKey": "${ANTHROPIC_API_KEY:-fallback-key}"
      }
    }
  }
}

Handling in Code

// Check if string contains env var reference
function containsEnvVarReference(value: string): boolean {
  return /\$\{[^}]+\}/.test(value);
}

// Collect all env var paths in an object
function collectEnvRefPaths(
  value: unknown,
  path: string,
  output: Map<string, string>
): void {
  if (typeof value === "string") {
    if (containsEnvVarReference(value)) {
      output.set(path, value);
    }
    return;
  }
  if (Array.isArray(value)) {
    value.forEach((item, index) => {
      collectEnvRefPaths(item, `${path}[${index}]`, output);
    });
    return;
  }
  if (isPlainObject(value)) {
    for (const [key, child] of Object.entries(value)) {
      const childPath = path ? `${path}.${key}` : key;
      collectEnvRefPaths(child, childPath, output);
    }
  }
}

Schema Validation

Zod Schema Pattern

OpenClaw uses Zod for runtime validation:

import { z } from "zod";

// Define schema
const AgentConfigSchema = z.object({
  model: z.string().optional(),
  temperature: z.number().min(0).max(2).optional(),
  maxTokens: z.number().positive().optional(),
  enabled: z.boolean().default(true),
});

// Validate
type AgentConfig = z.infer<typeof AgentConfigSchema>;

function validateConfig(data: unknown): AgentConfig {
  return AgentConfigSchema.parse(data);
}

// Safe validation
function safeValidateConfig(data: unknown): AgentConfig | null {
  const result = AgentConfigSchema.safeParse(data);
  return result.success ? result.data : null;
}

Common OpenClaw Schema Types

// Model reference: "provider/model-name"
const ModelRefSchema = z.string().regex(/^[a-z0-9-]+\/[a-z0-9-]+$/i);

// Channel ID
const ChannelIdSchema = z.enum([
  "telegram", "discord", "slack", "whatsapp",
  "signal", "imessage", "irc", "web"
]);

// Duration string: "30s", "5m", "1h"
const DurationSchema = z.string().regex(/^\d+[smhd]$/);

Config Includes

OpenClaw supports config file includes:

{
  "include": [
    "./base-config.json",
    "~/.openclaw/shared-channels.json"
  ],
  "agents": {
    // Local overrides
  }
}

Processing Order

1. Load included files (recursive, depth-limited)
2. Merge in order (later files override earlier)
3. Apply env var substitution
4. Validate against schema
5. Apply runtime overrides

jq Patterns for OpenClaw

Common Operations

# Pretty print OpenClaw config
jq . ~/.openclaw/config.json

# Get gateway mode
jq '.gateway.mode' ~/.openclaw/config.json

# List all agent IDs
jq '.agents | keys[]' ~/.openclaw/config.json

# Find agent using specific model
jq '.agents | to_entries[] | select(.value.model == "anthropic/claude-opus-4-6") | .key' ~/.openclaw/config.json

# Get all channel types
jq '.channels | keys[]' ~/.openclaw/config.json

# Check if Telegram is configured
jq '.channels.telegram != null' ~/.openclaw/config.json

# Extract all model references
jq '.. | objects | select(has("model")) | .model' ~/.openclaw/config.json

# Merge patch using jq
jq '.agents.main.model = "anthropic/claude-opus-4-6"' ~/.openclaw/config.json > tmp.json \
  && mv tmp.json ~/.openclaw/config.json

Advanced jq

# Deep search for all API keys (for audit)
jq '.. | objects | .apiKey? // .token? // .password? | select(.)' ~/.openclaw/config.json

# Collect all environment variable references
jq -r '.. | strings | select(contains("${"))' ~/.openclaw/config.json

# Validate JSON structure (returns true/false)
jq 'if has("gateway") and has("agents") then true else false end' ~/.openclaw/config.json

# Create minimal config from full config
jq '{ gateway: .gateway, agents: { main: .agents.main } }' ~/.openclaw/config.json

Common Config Patterns

Gateway Configuration

{
  "gateway": {
    "mode": "http",  // "http", "disabled", "process"
    "http": {
      "bind": "127.0.0.1",
      "port": 3000,
    },
    "auth": {
      "token": "${OPENCLAW_GATEWAY_TOKEN}",
    },
  },
}

Agent Configuration

{
  "agents": {
    "main": {
      "model": "anthropic/claude-opus-4-6",
      "temperature": 0.7,
      "maxTokens": 4096,
      // System prompt or reference to file
      "systemPrompt": "You are a helpful assistant.",
      "systemPromptFile": "~/.openclaw/agents/main/prompt.md",
    },
    "coder": {
      "model": "anthropic/claude-sonnet-4-5",
      "temperature": 0.2,
      // Inherit from main with overrides
      "inherits": "main",
    },
  },
}

Channel Configuration

{
  "channels": {
    "telegram": {
      "botToken": "${TELEGRAM_BOT_TOKEN}",
      "allowFrom": ["@username"],
    },
    "discord": {
      "botToken": "${DISCORD_BOT_TOKEN}",
      "applicationId": "123456789",
    },
    "slack": {
      "botToken": "${SLACK_BOT_TOKEN}",
      "appToken": "${SLACK_APP_TOKEN}",
    },
  },
}

Tools Configuration

{
  "tools": {
    "alsoAllow": ["web_search", "browser"],
    "deny": ["exec"],
    "config": {
      "web_search": {
        "provider": "brave",
        "apiKey": "${BRAVE_API_KEY}",
      },
    },
  },
}

Validation & Error Handling

Common Validation Errors

// Schema validation errors provide detailed paths
const result = schema.safeParse(data);
if (!result.success) {
  for (const error of result.error.errors) {
    console.log(`${error.path.join('.')}: ${error.message}`);
    // e.g., "agents.main.temperature: Number must be less than or equal to 2"
  }
}

Config File Recovery

# If config is corrupted, OpenClaw keeps backups
ls -la ~/.openclaw/config.json.*

# Restore from backup
cp ~/.openclaw/config.json.2024-01-15T10-30-00.bak ~/.openclaw/config.json

# Or use OpenClaw's built-in rotation
openclaw config restore

Best Practices

1. Always Validate After Edit

# Validate config syntax and schema
openclaw config validate

# Test config loading
openclaw config get

2. Backup Before Changes

# Create timestamped backup
cp ~/.openclaw/config.json ~/.openclaw/config.json.$(date +%Y%m%d_%H%M%S).bak

3. Use Type Guards

// Never assume structure - always validate
if (!isPlainObject(config.agents)) {
  throw new Error("Invalid agents configuration");
}

4. Handle Env Vars Carefully

// Preserve env var references when editing
const originalValue = "${API_KEY}";
const newValue = process.env.API_KEY || originalValue;

5. Use Structured Clone for Deep Copies

// Preferred for deep cloning
deepCopy = structuredClone(original);

// Fallback for older environments
deepCopy = JSON.parse(JSON.stringify(original));

6. Atomic Writes

// Write to temp file, then rename
fs.writeFileSync(tempPath, data);
fs.renameSync(tempPath, finalPath);

Security Considerations

• File permissions: Config files should be 0o600 (user read/write only)
• No secrets in JSON: Use ${ENV_VAR} substitution
• Validate inputs: Always schema-validate external JSON
• Sanitize paths: Use path.resolve() and check traversal
• Audit logging: OpenClaw logs config changes to config-audit.jsonl

Troubleshooting

Common Issues

Debug Commands

# Check raw config (before env substitution)
cat ~/.openclaw/config.json

# Check effective config (after all processing)
openclaw config get --json

# List all env var references
openclaw config env-refs

# Trace config loading
OPENCLAW_DEBUG=config openclaw config get

Editing Providers & Model Configuration

When adding or updating AI providers in openclaw.config.json, you must discover actual model names from the provider's API and handle reasoning model variants correctly.

Model Discovery Workflow

# 1. Fetch available models from provider API
# xAI example - requires XAI_API_KEY
XAI_API_KEY="your-key"
curl -s -H "Authorization: Bearer $XAI_API_KEY" \
  https://api.x.ai/v1/models | jq '.data[] | {id: .id, name: .object}'

# OpenAI example
curl -s -H "Authorization: Bearer $OPENAI_API_KEY" \
  https://api.openai.com/v1/models | jq '.data[] | select(.id | contains("gpt")) | .id'

# Together AI example
curl -s -H "Authorization: Bearer $TOGETHER_API_KEY" \
  https://api.together.xyz/v1/models | jq '.[] | {id: .id, name: .display_name}'

Provider Configuration Schema

OpenClaw uses ModelProviderConfig schema:

type ModelProviderConfig = {
  baseUrl: string;           // API endpoint base URL
  apiKey?: string;           // Optional: API key (prefer env vars)
  auth?: "api-key" | "aws-sdk" | "oauth" | "token";
  api?: "openai-completions" | "openai-responses" | 
        "anthropic-messages" | "google-generative-ai" |
        "github-copilot" | "bedrock-converse-stream" | "ollama";
  headers?: Record<string, string>;  // Custom headers
  models: ModelDefinitionConfig[];   // Model definitions
};

type ModelDefinitionConfig = {
  id: string;                // Model ID (e.g., "grok-4")
  name: string;              // Display name (e.g., "Grok 4")
  api?: ModelApi;            // Override API type per model
  reasoning: boolean;        // Whether model supports reasoning/thinking
  input: Array<"text" | "image">;  // Supported input types
  cost: {
    input: number;           // Cost per 1M input tokens
    output: number;          // Cost per 1M output tokens
    cacheRead: number;       // Cost per 1M cached tokens read
    cacheWrite: number;      // Cost per 1M cached tokens written
  };
  contextWindow: number;     // Max context window size
  maxTokens: number;         // Max output tokens
  headers?: Record<string, string>;
  compat?: ModelCompatConfig;
};

Reasoning Model Families

CRITICAL: Some models have reasoning variants handled specially by OpenClaw. For example, xAI's grok-4-1-fast has three variants:

In OpenClaw, you typically configure ONLY the base model (grok-4-1-fast). The system automatically switches between reasoning/non-reasoning variants based on the thinking directive or configuration.

{
  "models": {
    "providers": {
      "xai": {
        "baseUrl": "https://api.x.ai/v1",
        "api": "openai-completions",
        "apiKey": "${XAI_API_KEY}",
        "models": [
          {
            "id": "grok-4-1-fast",
            "name": "Grok 4.1 Fast",
            "reasoning": false,  // Base model is non-reasoning
            "input": ["text"],
            "cost": { "input": 0, "output": 0, "cacheRead": 0, "cacheWrite": 0 },
            "contextWindow": 128000,
            "maxTokens": 8192
          }
          // NOTE: Do NOT add -reasoning or -non-reasoning variants separately
          // OpenClaw handles these automatically via model family resolution
        ]
      }
    }
  }
}

Model Family Resolution

OpenClaw internally defines reasoning model families in src/agents/model-families.ts:

const REASONING_MODEL_FAMILIES = [
  {
    provider: "xai",
    members: [
      "grok-4-1-fast",
      "grok-4-1-fast-reasoning",
      "grok-4-1-fast-non-reasoning"
    ],
    reasoningModel: "grok-4-1-fast-reasoning",
    nonReasoningModel: "grok-4-1-fast-non-reasoning",
  },
];

When a user requests a model with thinking: "on" or thinking: "off", OpenClaw:

1. Looks up if the requested model is in a reasoning family
2. If thinking: "on" → uses reasoningModel variant
3. If thinking: "off" → uses nonReasoningModel variant
4. If no thinking directive → uses the base model as-is

Complete Provider Configuration Example

{
  "models": {
    "mode": "merge",  // "merge" or "replace"
    "providers": {
      // xAI - Grok models with reasoning variants
      "xai": {
        "baseUrl": "https://api.x.ai/v1",
        "api": "openai-completions",
        "apiKey": "${XAI_API_KEY}",
        "models": [
          {
            "id": "grok-4-1-fast",
            "name": "Grok 4.1 Fast",
            "reasoning": false,
            "input": ["text"],
            "cost": { "input": 0, "output": 0, "cacheRead": 0, "cacheWrite": 0 },
            "contextWindow": 128000,
            "maxTokens": 8192
          },
          {
            "id": "grok-4",
            "name": "Grok 4",
            "reasoning": false,
            "input": ["text", "image"],  // Vision-capable
            "cost": { "input": 0, "output": 0, "cacheRead": 0, "cacheWrite": 0 },
            "contextWindow": 128000,
            "maxTokens": 8192,
            "compat": {
              "supportsReasoningEffort": false,
              "maxTokensField": "max_completion_tokens"
            }
          }
        ]
      },
      
      // OpenAI - with response API and reasoning
      "openai": {
        "baseUrl": "https://api.openai.com/v1",
        "api": "openai-responses",
        "apiKey": "${OPENAI_API_KEY}",
        "models": [
          {
            "id": "gpt-5.2",
            "name": "GPT-5.2",
            "reasoning": false,
            "input": ["text", "image"],
            "cost": { "input": 2.5, "output": 10, "cacheRead": 0.5, "cacheWrite": 1.25 },
            "contextWindow": 200000,
            "maxTokens": 16384,
            "compat": {
              "supportsReasoningEffort": true,
              "thinkingFormat": "openai"
            }
          },
          {
            "id": "o3-mini",
            "name": "o3 Mini",
            "reasoning": true,  // Built-in reasoning model
            "input": ["text", "image"],
            "cost": { "input": 1.1, "output": 4.4, "cacheRead": 0.275, "cacheWrite": 0.55 },
            "contextWindow": 200000,
            "maxTokens": 100000,
            "compat": {
              "supportsReasoningEffort": true,
              "requiresAssistantAfterToolResult": true
            }
          }
        ]
      },
      
      // Anthropic - Messages API
      "anthropic": {
        "baseUrl": "https://api.anthropic.com",
        "api": "anthropic-messages",
        "apiKey": "${ANTHROPIC_API_KEY}",
        "models": [
          {
            "id": "claude-opus-4-6",
            "name": "Claude Opus 4.6",
            "reasoning": false,
            "input": ["text", "image"],
            "cost": { "input": 15, "output": 75, "cacheRead": 1.88, "cacheWrite": 7.5 },
            "contextWindow": 200000,
            "maxTokens": 8192,
            "compat": {
              "supportsStore": false,
              "supportsDeveloperRole": false
            }
          }
        ]
      },
      
      // Google Gemini
      "google": {
        "baseUrl": "https://generativelanguage.googleapis.com/v1beta",
        "api": "google-generative-ai",
        "apiKey": "${GEMINI_API_KEY}",
        "models": [
          {
            "id": "gemini-3-pro-preview",
            "name": "Gemini 3 Pro Preview",
            "reasoning": false,
            "input": ["text", "image"],
            "cost": { "input": 1.25, "output": 10, "cacheRead": 0.31, "cacheWrite": 1.25 },
            "contextWindow": 1000000,
            "maxTokens": 8192,
            "compat": {
              "thinkingFormat": "qwen"
            }
          }
        ]
      },
      
      // Ollama - local models (auto-discovered)
      "ollama": {
        "baseUrl": "http://localhost:11434/v1",
        "api": "ollama",
        "models": []  // Auto-populated from /api/tags
      }
    }
  }
}

Model Compatibility Flags

type ModelCompatConfig = {
  // OpenAI-specific features
  supportsStore?: boolean;                    // Use 'store' parameter
  supportsDeveloperRole?: boolean;            // Use 'developer' vs 'system' role
  supportsReasoningEffort?: boolean;          // Support reasoning_effort param
  supportsUsageInStreaming?: boolean;         // Usage in streaming responses
  supportsStrictMode?: boolean;               // Strict tool mode
  
  // Token handling
  maxTokensField?: "max_completion_tokens" | "max_tokens";
  
  // Thinking/reasoning format
  thinkingFormat?: "openai" | "zai" | "qwen";
  
  // Tool calling quirks
  requiresToolResultName?: boolean;           // Must include tool result name
  requiresAssistantAfterToolResult?: boolean; // Assistant message after tool
  requiresThinkingAsText?: boolean;           // Thinking blocks as text
  requiresMistralToolIds?: boolean;           // Mistral-style tool IDs
};

Validating Provider Configuration

# Validate the full config including models
openclaw config validate

# Check if models.json is correctly generated
openclaw models list

# Test a specific model provider
openclaw models test --provider xai --model grok-4-1-fast

# Debug model resolution
OPENCLAW_DEBUG=models openclaw models list

Common Pitfalls

Provider-Specific Notes

xAI (Grok)

• Uses openai-completions API
• Model families auto-resolve reasoning variants
• Vision support varies by model

OpenAI

• Use openai-responses for o-series and GPT-5
• Use openai-completions for legacy GPT-4
• Reasoning effort adjustable via supportsReasoningEffort

Anthropic

• Uses anthropic-messages API
• No separate reasoning models (all models can think)
• Distinct cost structure for prompt caching

Google (Gemini)

• Uses google-generative-ai API
• Very large context windows (1M tokens)
• Different content format than OpenAI/Anthropic

Ollama

• Set api: "ollama" for native discovery
• Models auto-discovered from /api/tags
• Local inference - no API key needed

Model Aliases

Define aliases for common models in agent defaults:

{
  "agents": {
    "defaults": {
      "models": {
        "fast": { "alias": "Grok Fast", "id": "xai/grok-4-1-fast" },
        "smart": { "alias": "Claude Opus", "id": "anthropic/claude-opus-4-6" },
        "vision": { "alias": "GPT Vision", "id": "openai/gpt-5.2" }
      }
    }
  }
}

Use aliases in agent config:

{
  "agents": {
    "main": {
      "model": "fast"  // Resolves to xai/grok-4-1-fast
    }
  }
}