Agent Configuration¶

Configure the primary coding agent for the workspace: provider, model, system prompt, tools, and MCP servers.

Configuration¶

[workspace.agent]
provider = "claude-code"
model = "claude-sonnet-4-20250514"
tools_enabled = true
system_prompt = """
You are a senior developer working on this project.
Follow the coding standards and write tests for all changes.
"""

[[workspace.agent.mcp_servers]]
name = "filesystem"
command = "npx"
args = ["-y", "@anthropic-ai/mcp-server-filesystem", "/workspace"]

Fields¶

Field	Type	Required	Default	Description
`provider`	string	Yes	—	Agent provider: `claude-code`, `codex`, `gemini`, `cursor`
`model`	string	No	provider default	Model identifier
`system_prompt`	string	No	—	System prompt prepended to all sessions
`max_tokens`	integer	No	—	Max tokens per response
`temperature`	float	No	—	Sampling temperature
`tools_enabled`	boolean	No	`true`	Whether the agent can use tools
`mcp_servers`	array	No	`[]`	MCP server configurations
`allowed_tools`	string[]	No	`[]`	Tool allowlist (empty = all)
`denied_tools`	string[]	No	`[]`	Tool denylist
`extra`	object	No	`{}`	Provider-specific settings

Providers¶

Provider	Description
`claude-code`	Anthropic's Claude Code CLI
`codex`	OpenAI Codex agent
`gemini`	Google Gemini CLI
`cursor`	Cursor CLI agent

System Prompts¶

System prompts give the agent context about the workspace:

[workspace.agent]
provider = "claude-code"
system_prompt = """
Project: E-commerce Platform
Stack: React frontend (/workspace/frontend), Rust backend (/workspace/backend)
Database: PostgreSQL
Standards: Use TypeScript strict mode, write unit tests, follow REST conventions.
"""

MCP Servers¶

Connect MCP (Model Context Protocol) servers for extended capabilities:

[[workspace.agent.mcp_servers]]
name = "filesystem"
command = "npx"
args = ["-y", "@anthropic-ai/mcp-server-filesystem", "/workspace"]

[[workspace.agent.mcp_servers]]
name = "postgres"
command = "npx"
args = ["-y", "@anthropic-ai/mcp-server-postgres", "postgresql://localhost/mydb"]