> ## Documentation Index
> Fetch the complete documentation index at: https://docs.llmeasy.ru/llms.txt
> Use this file to discover all available pages before exploring further.

# Best practices

> Tips for getting the most out of LLMEasy.

## Keep your API Key secure

**Use environment variables — never hardcode.** Writing your API Key directly into source code or config files risks accidental exposure in version control. Recommended approach:

```bash theme={null}
# Add to ~/.zshrc or ~/.bashrc so it loads automatically each session
export LLMEASY_API_KEY="your-api-key-here"
```

For Claude Code, manage secrets through the `env` field in `settings.json` rather than exposing them in your shell config:

```json theme={null}
{
  "env": {
    "ANTHROPIC_AUTH_TOKEN": "your-api-key-here"
  }
}
```

**Rotate your API Key periodically.** If you suspect exposure, delete the key immediately from the dashboard and generate a new one.

***

## Set a generous timeout

AI model responses can take time, especially for complex reasoning tasks. Set the timeout high enough to avoid premature disconnects:

```json theme={null}
{
  "env": {
    "API_TIMEOUT_MS": "3000000"
  }
}
```

`3000000` ms = 50 minutes. This is appropriate for deep reasoning or long agentic tasks.

***

## Clear conflicting environment variables

If you've previously used the official Anthropic or OpenAI APIs — or another relay service — stale environment variables may override your LLMEasy config and silently send requests to the wrong endpoint.

Before configuring LLMEasy, check and clear these:

```bash theme={null}
# Check if they exist
echo $ANTHROPIC_AUTH_TOKEN
echo $ANTHROPIC_BASE_URL
echo $OPENAI_API_KEY
echo $OPENAI_BASE_URL

# Clear them
unset ANTHROPIC_AUTH_TOKEN
unset ANTHROPIC_BASE_URL
unset OPENAI_API_KEY
unset OPENAI_BASE_URL
```

***

## Claude Code: check the Key group first

With the **Claude Key group**, **do not set** `ANTHROPIC_MODEL`, `ANTHROPIC_SMALL_FAST_MODEL`, `ANTHROPIC_DEFAULT_SONNET_MODEL`, or other model-specific variables. With the **GPT Key group**, add model mappings through `ANTHROPIC_DEFAULT_HAIKU_MODEL`, `ANTHROPIC_DEFAULT_SONNET_MODEL`, and `ANTHROPIC_DEFAULT_OPUS_MODEL`.

Do not mix the two setups, or requests may be routed to the wrong model.

**Disable non-essential traffic.** Setting `CLAUDE_CODE_DISABLE_NONESSENTIAL_TRAFFIC=1` reduces background requests from Claude Code, reserving your API quota for actual coding tasks.

**Do not add `/v1` to Claude Code.** `ANTHROPIC_BASE_URL` should be `https://www.llmeasy.ru`.

***

## Use one API Key across multiple tools

The same API Key works in Claude Code, Codex, and external tools. Quota is shared, which makes central management simpler.

But the endpoint must match the protocol:

* Claude Code / `ANTHROPIC_BASE_URL`: `https://www.llmeasy.ru`
* Codex / `OPENAI_BASE_URL` / external tools: `https://www.llmeasy.ru/v1`

If you need to track usage separately per project or team, create multiple API Keys in the dashboard and assign one to each tool or project.

For external tools, copy the model ID from the **GPT Key group** in the <a href={"https://www.llmeasy.ru/pricing"}>model plaza</a>.

***

## Troubleshooting checklist

1. **Confirm the API Key is correct** — check the dashboard, paste carefully with no extra spaces
2. **Confirm the endpoint for the protocol** — Claude Code uses `https://www.llmeasy.ru`; Codex / external tools use `https://www.llmeasy.ru/v1`
3. **Check for conflicting env vars** — run `echo $ANTHROPIC_AUTH_TOKEN`, `echo $OPENAI_BASE_URL`, and related commands
4. **Check the tool's official setup requirements** — for external tools, verify the `Base URL`, API key, model ID, and tool version first; only add extra headers when the tool's official docs or your upstream gateway explicitly requires them
5. **Check your balance** — requests are rejected when the balance is depleted
6. **Contact support** — if the above steps don't resolve the issue