Antigravity AI Agent Not Responding? Common Trouble FAQ

Setup and context

Using Antigravity's AI agent features can sometimes result in unresponsive agents, interrupted code generation, multi-agent failures, and other issues. This beginner-friendly FAQ covers 7 common agent-related problems and their solutions in Q&A format.

:::info This FAQ is based on the official Antigravity Agent API documentation and actual user support cases. :::

Q1: Agent Unresponsive / Timeout

Symptoms

• Running an agent command produces no output
• "Agent request timed out" error appears
• Nothing displays on screen after waiting several minutes

Root Causes and Solutions

Cause 1: Default Timeout is Too Short

Antigravity's default timeout is 30 seconds. Complex operations may exceed this limit.

Increase the timeout in your Antigravity config file (~/.config/antigravity/config.json):

{
  "agent": {
    "requestTimeout": 120000,  // Increase to 120 seconds
    "retries": 3
  }
}

Restart Antigravity afterward:

# Kill existing process
pkill -f claude-code
 
# Restart
claude-code

Cause 2: API Rate Limit Reached

Your Claude API or integrated service may have hit rate limits.

# Check API usage
curl -H "Authorization: Bearer $ANTHROPIC_API_KEY" \
  https://api.anthropic.com/v1/account/info
 
# Or check Antigravity logs
tail -f ~/.config/antigravity/logs/agent.log | grep -i "rate"

If rate-limited, try these solutions:

1. Wait a few minutes before retrying
2. Verify ANTHROPIC_API_KEY is set correctly
3. Consider upgrading your API plan

Cause 3: Unstable Network Connection

Network disruption during agent execution causes timeouts.

# Check network connectivity
ping -c 5 api.anthropic.com
 
# Verify DNS resolution
nslookup api.anthropic.com
 
# If on WiFi, try wired connection or reset WiFi

Enable retry logic for network instability:

{
  "agent": {
    "networkRetry": {
      "enabled": true,
      "maxRetries": 5,
      "delayMs": 2000
    }
  }
}

Cause 4: System Resources Exhausted

Memory or CPU exhaustion causes agent processing delays.

# Check resource usage
# macOS/Linux
top -b -n 1
 
# Windows (PowerShell)
Get-Process | Sort-Object WorkingSet -Descending | Select-Object -First 10

Close unnecessary processes and increase Node.js heap size:

NODE_OPTIONS="--max-old-space-size=4096" claude-code

:::tip Large operations consume more CPU and memory. Close extra browser tabs and applications to free up resources before execution. :::

Q2: Code Generation Stops Midway

Symptoms

• Agent starts generating code but stops partway through
• "Generation interrupted" error appears
• Only partial code is generated

Root Causes and Solutions

Cause 1: Reached Token Limit

Claude's context window has a limited number of available tokens. Processing large files can exhaust this limit.

Check current token usage in logs:

# Check Antigravity logs
tail -f ~/.config/antigravity/logs/agent.log | grep -i "token"

If token limit is reached, try these solutions:

1. Split files: Divide large files into smaller, more manageable pieces
2. Clear history: Remove previous conversation history
3. Simplify prompts: Use shorter, more specific instructions

# Clear conversation history
# UI: Menu → Clear Conversation History

Cause 2: Memory Exhaustion During Processing

Analyzing large projects requires significant memory.

# Check memory usage
# macOS
ps aux | grep claude-code | grep -v grep
 
# Increase memory and restart
NODE_OPTIONS="--max-old-space-size=8192" claude-code

Cause 3: Agent Context Management Error

If referenced files are deleted or moved during processing, generation stops.

# Verify files still exist during processing
ls -la src/
ls -la tests/

Cause 4: Plugin or Integration Failure

External plugins (linters, formatters) failing causes generation interruption.

# List installed skills
claude skills list
 
# Uninstall problematic skill
claude uninstall skills.sh/skill-name

:::warning Don't manually edit project files while code generation is running. This confuses the agent and may interrupt generation. :::

Q3: Agent Edits Wrong File

Symptoms

• Agent modifies files you didn't specify
• File in src/ gets edited in lib/ instead
• File paths are misinterpreted

Root Causes and Solutions

Cause 1: Ambiguous File Paths

When multiple files share the same name, the agent may pick the wrong one.

# Find duplicate filenames
find . -name "component.tsx" -type f
 
# Example output:
# ./src/components/component.tsx
# ./lib/components/component.tsx

Always use complete file paths in instructions:

"Edit ./src/components/Header.tsx"

Cause 2: Relative Path Interpretation Mismatch

Antigravity and the actual filesystem may interpret relative paths differently.

# Check current working directory
pwd
 
# Check Antigravity logs
tail -f ~/.config/antigravity/logs/agent.log | grep -i "working directory"

Always use absolute paths or relative paths from project root:

// Good
"$PROJECT_ROOT/src/utils/helper.ts"

// Avoid
"../../../utils/helper.ts"  // Ambiguous

Cause 3: Agent Scope Not Restricted

If you want to limit agent workspace, specify allowed directories:

{
  "agent": {
    "workspace": {
      "allowedDirs": [
        "src/",
        "lib/",
        "tests/"
      ],
      "excludeDirs": [
        "node_modules/",
        ".git/",
        "build/"
      ]
    }
  }
}

Add this to ~/.config/antigravity/config.json.

Cause 4: File Index Cache is Stale

Large projects may have outdated file mappings in cache.

# Clear cache
rm -rf ~/.config/antigravity/cache/file-index
 
# Restart Antigravity
pkill -f claude-code
claude-code

:::info Antigravity creates a file index on first startup. Large projects may take several minutes. :::

Q4: Multi-Agent Processing Doesn't Run Concurrently

Symptoms

• Multiple agent processes run sequentially instead of in parallel
• No concurrent processing occurs
• Performance is unexpectedly slow

Root Causes and Solutions

Cause 1: Multi-Agent Feature Disabled

Multi-agent processing may be disabled by default. Enable it in config:

{
  "agent": {
    "multiAgent": {
      "enabled": true,
      "maxConcurrent": 4,
      "queueStrategy": "priority"
    }
  }
}

Increase maxConcurrent to allow more simultaneous agents.

Cause 2: Insufficient System Memory

Multiple concurrent agents consume significant memory.

# Check memory availability
free -h  # Linux
vm_stat  # macOS

If low on memory:

1. Close unnecessary applications
2. Reduce maxConcurrent (e.g., 4 → 2)
3. Increase Node.js heap size

NODE_OPTIONS="--max-old-space-size=6144" claude-code

Cause 3: API Rate Limiting Enforces Sequential Execution

Claude API rate limits may force sequential execution despite settings.

# Check API status
curl -s -H "Authorization: Bearer $ANTHROPIC_API_KEY" \
  https://api.anthropic.com/v1/account/info | jq .
 
# Check logs
tail -f ~/.config/antigravity/logs/agent.log | grep -i "concurrent\|rate"

Cause 4: Unresolved Dependencies Between Agents

When multiple agents target the same file, Antigravity auto-switches to sequential execution.

# Check for dependency issues in logs
tail -f ~/.config/antigravity/logs/agent.log | grep -i "dependency\|lock"

If dependencies exist, explicitly specify execution order.

:::tip When running multi-agent processes, adding slight delays between agents improves stability. :::

Q5: Context Too Long, Agent Gets Confused

Symptoms

• Agent misses important details when processing large files
• Agent output is illogical or contradictory
• "Context too long" warning appears

Root Causes and Solutions

Cause 1: Files Too Large for Context

Agent context window has limits (Claude 3 Opus: 200K tokens). Split large files:

# Check file size
wc -l src/main.ts
 
# Split large file into chunks
split -l 500 src/main.ts src/main_part_
 
# Or use specialized tools
csplit src/main.ts '/export/ {*}'

Cause 2: Unnecessary Information in Context

Verbose prompts dilute important instructions.

Keep prompts concise:

# Bad (too long)
"Look at src/main.ts, create handleSubmit in LoginComponent.
By the way, this project uses React and TypeScript and..."

# Good (concise)
"Add handleSubmit to LoginComponent in src/main.ts.
Requirements: form validation then API call"

Cause 3: Conversation History Accumulation

Extended conversations fill context window quickly. Periodically reset:

# Use UI "Clear Conversation" option
# Or CLI
claude clear-conversation

Cause 4: Too Many Reference Files

Passing multiple files at once exhausts context. Use only necessary files:

# Bad
"View all files in src/components/*.tsx and src/utils/*.ts..."

# Good
"Reference src/components/Button.tsx and src/utils/validator.ts..."

:::warning When context is near capacity, agent output quality degrades significantly. Keep context usage below 70% by splitting files and conversations. :::

Q6: Agent Responds in English Only

Symptoms

• Agent responds in English despite Japanese instructions
• Code comments appear in English
• Explanations are only in English

Root Causes and Solutions

Cause 1: Antigravity Default Language Set to English

Check and change Antigravity's language setting:

// ~/.config/antigravity/config.json
{
  "language": "en",
  "agent": {
    "systemPrompt": "You are a helpful assistant. Always respond in the user's language."
  }
}

Restart Antigravity:

pkill -f claude-code
claude-code

Cause 2: Agent System Prompt in English Only

Update the agent system prompt to support your language:

{
  "agent": {
    "systemPrompt": "You are a helpful assistant. Always respond in the user's language."
  }
}

Cause 3: Language Model Optimized for English

Some Claude models lean toward English. Explicitly request your language:

Example:
"Create the following component. Respond entirely in Japanese.
Requirements: Button component, TypeScript, React"

Cause 4: Mixed Languages Causing Default to English

Mixed language usage defaults to English. Use one language consistently.

:::info Claude supports multiple languages, but explicit requests improve reliability. :::

Q7: Unexpectedly High API Usage

Symptoms

• Claude API bill spikes suddenly
• Large number of API calls in short time
• Monthly usage exceeds budget

Root Causes and Solutions

Cause 1: Debug Mode Enabled

Debug mode creates extensive logging, causing many API calls. Disable for production:

{
  "debug": false,
  "agent": {
    "verbose": false,
    "logAllRequests": false
  }
}

Cause 2: Excessive Retry Logic

Frequent errors trigger retries, increasing API calls.

{
  "agent": {
    "requestRetry": {
      "maxRetries": 2,      // Default: 3
      "delayMs": 1000
    }
  }
}

Cause 3: Repeated High-Token Operations

Repeatedly analyzing or refactoring large files increases token usage. Use caching:

# Save and reuse agent results to reduce repeated calls
# Cache agent output for similar requests

Cause 4: Simultaneous Multi-Agent Execution

Parallel agents significantly increase API calls.

{
  "agent": {
    "multiAgent": {
      "maxConcurrent": 1  // Limit to sequential
    }
  }
}

Cause 5: Unnecessary API-Dependent Features Enabled

Verify no unneeded integrations are active:

# List active integrations
claude integrations list
 
# Disable unused ones
claude integrations disable integration-name

:::tip Monitor API usage regularly:

• Check Claude API dashboard statistics
• Review Antigravity logs for API calls
• Track monthly usage trends :::

Summary

Most Antigravity agent issues are resolved with these approaches:

If issues persist, check logs and contact support:

# View agent logs
cat ~/.config/antigravity/logs/agent.log | tail -100