Skip to main content
Tembo supports multiple coding agents to solve your software engineering tasks.

​
Available Coding Agents

Coding AgentModel ProvidersDefault Model
Claude CodeAnthropicclaude-opus-4-5
CodexOpenAIgpt-5-medium
OpencodeAnthropicclaude-opus-4-5
AmpAnthropic (auto-selected)claude-opus-4-5
CursorAnthropic/OpenAI/Google/Cursor/XAIclaude-opus-4-5

​
Claude Code (Default)

Powered by Anthropic’s Claude models. Best balance of quality, reliability, and cost. Models: claude-opus-4-5 (default), claude-4-5-sonnet, claude-4.1-opus, claude-4-5-haiku, claude-4-sonnet

​
Codex

Powered by OpenAI’s GPT-5 models with configurable reasoning levels. Models: gpt-5-medium (default), gpt-5-minimal, gpt-5-low, gpt-5-high, gpt-5-codex

​
Opencode

Alternative agent supporting Anthropic Claude models. Models: claude-opus-4-5 (default), claude-4-5-sonnet, claude-4.1-opus, claude-4-5-haiku, claude-4-sonnet

​
Amp

Powered by Claude models. Manages model selection automatically, defaults to claude-opus-4-5.

​
Cursor

Supports Claude, GPT, Gemini, Composer, and Grok through the Cursor CLI. Models: claude-opus-4-5 (default), claude-4-5-sonnet, claude-4.1-opus, gpt-5.1, gpt-5.1-codex, gpt-5.1-codex-high, gpt-5.2, gemini-3-pro, composer-1, grok

​
Selecting an Agent

Use the format agentType:model: 
claudeCode:claude-4-5-sonnet
codex:gpt-5-high
cursor:grok
Default: claudeCode:claude-opus-4-5

​
Choosing the Right Model

Use CaseRecommended Models
Most tasksclaude-4-5-sonnet, gpt-5-medium
Quick fixesclaude-4-5-haiku, gpt-5-minimal
Complex refactoringclaude-opus-4-5, gpt-5-high

​
Rule Files

All agents use rule files (AGENTS.md, tembo.md, .cursorrules, etc.) to understand your coding standards. See Rule Files for details.

​
Configuration

Set your coding agent via:
  • Dashboard Settings — Default for all tasks
  • API — Specify per-task with agent key
  • Integration Settings — Per-integration preferences

​
Troubleshooting

Task taking too long? Break into smaller tasks, or use a faster model (Sonnet/Haiku instead of Opus). Unexpected output? Check your rule files, add more context to task descriptions, or use the Feedback Loop via PR comments. Model unavailable? Verify the model ID or omit it to use the default. Contact [email protected] if issues persist.