™
Dashboard →
Which Coding Tools Does GetWrangler Support?
An honest, up-to-date breakdown of what's routable through GetWrangler today, what's partial, and what isn't supported yet.
GetWrangler is the AI Control Tower for Spend, Routing, and Governance. For AI coding tools, it helps teams attribute usage, manage chargeback, and understand who is consuming AI resources across the organization.
Coding-tool tracking is the visibility layer. Production AI spend governance is where GetWrangler scales.
Compatibility
Coding Tool Compatibility
This table reflects current, confirmed behavior — not aspirational roadmap status.
| Tool | Current Support | What's Tracked | Token / Cost Savings Available | Status |
|---|---|---|---|---|
| GitHub Copilot CLI | Full support | Chat, streaming, and tool-calling / agentic requests | Yes | Available Now |
| Cursor | Chat & Plan mode only | Chat / planning requests | Yes (chat/plan traffic only) | Available Now (Partial) |
| Claude Code | Tracking/billing only (direct to your own Anthropic account) | Chat, streaming, and tool-calling/agentic requests | No — full capability preserved, no substitution | Available Now (Tracking Only) |
| Codex CLI (OpenAI) | Tracking/billing only (direct to your own OpenAI account) | Chat, streaming, and tool-calling/agentic requests | No — full capability preserved, no substitution | Available Now (Tracking Only) |
Why the difference
Why Some Tools Aren't Fully Supported Yet
Full support today, including real agentic and tool-use workflows — not just chat. Copilot CLI's default connection format lines up closely with what GetWrangler already speaks, so chat, streaming responses, and multi-turn tool-calling (where the CLI reads files or runs commands as part of a conversation) all route through GetWrangler correctly today.
Cursor lets you point its chat and plan panel at a custom OpenAI-compatible endpoint, which is the part GetWrangler supports today. Cursor's Composer/Agent mode, inline edit, and autocomplete are built to always call Cursor's own backend directly — this is a limitation of Cursor's own product design, not a configuration setting, so these modes cannot be routed through GetWrangler no matter how you configure your account.
Claude Code connects directly to your own Anthropic account, and Codex CLI connects directly to your own OpenAI account — both using their native request format rather than being routed and substituted the way GitHub Copilot CLI is today. This means every feature works exactly as it does natively — tool-calling, extended thinking, and anything else your account supports — because nothing about the request is being changed. The tradeoff is that tier-routing savings don't apply to this traffic today. An opt-in toggle to enable substitution-based savings for Claude Code and Codex CLI is planned but not yet built.
FAQ
Common Questions
Cursor's Composer/Agent mode, inline edit, and autocomplete are hardwired to call Cursor's own backend directly, regardless of any settings you configure — this is a limitation of Cursor's own product design, not something GetWrangler can route around. Only Cursor's chat/plan panel can be pointed at a custom OpenAI-compatible endpoint, which is the part GetWrangler supports today.
Claude Code and Codex CLI connect directly to your own Anthropic or OpenAI account using each vendor's native request format, rather than through GetWrangler's model-substitution routing. This is a deliberate design choice: it guarantees every feature — tool-calling, extended thinking, and anything else your account supports — works exactly as it would natively, since nothing about the request is being changed. The tradeoff is that today's setup doesn't apply tier-routing savings to that traffic; you get usage tracking, cost visibility, and billing attribution, but the request goes to the model you specified. We're planning an opt-in toggle that would let you trade some of that guaranteed-native behavior for tier-routing savings on this traffic, but it hasn't been built yet.
These are two different capabilities, and it's worth keeping them separate. Tracked means GetWrangler sees the request at all — it's logged with cost, token counts, and latency, giving you visibility and attribution even if you never touch routing. Cost savings means GetWrangler additionally applies its routing logic to reduce what that request costs.
For Cursor, only chat/plan traffic reaches GetWrangler at all, so tracking and any potential savings are limited to that traffic — Composer/Agent, inline edit, and autocomplete never reach us, so there's nothing to track or optimize there. For GitHub Copilot CLI, every request in the workflow — including tool-calling/agentic turns — reaches GetWrangler, so both tracking and cost-saving routing apply across the entire session, not just plain chat messages. For Claude Code and Codex CLI, every request reaches GetWrangler and is tracked in full, but no routing/substitution is applied — capability is fully preserved, and there's currently no savings on this traffic, by design.
Most teams running coding assistants through GetWrangler choose bypass mode, which preserves full model capability with zero substitution and shows $0 savings — the value there is attribution, not optimization. Cost-optimized routing is available for supported tools if you'd rather opt into it.