How to Chain Multiple Tools Together in Codex

Codex CLI is focused on code tasks, but those tasks often need real-world context — current library docs, live data, competitor analysis. AgentPatch gives Codex access to the full tool marketplace. When a task benefits from multiple tools, Codex uses them together.

Why This Matters

Terminal-first developers tend to want fewer context switches. If Codex can pull data, do research, and write code in the same session, that’s better than bouncing between a terminal, a browser, and whatever other tools you’re juggling.

Multi-tool capability matters most for tasks that would otherwise take several manual steps: research followed by code, data gathering followed by analysis, lookups followed by generation.

Setup

Add AgentPatch to ~/.codex/config.toml:

[mcp_servers.agentpatch]
url = "https://agentpatch.ai/mcp"
bearer_token_env_var = "AGENTPATCH_API_KEY"

Then set your API key:

export AGENTPATCH_API_KEY=your_api_key

Replace your_api_key with your actual key from the AgentPatch dashboard. Codex discovers all AgentPatch tools automatically on next start.

Example

You’re building a local search feature and want to test how it behaves when you combine place data with trend context. You tell Codex:

“For the city of Austin, search for coworking spaces, check if remote work is trending up or down, and pull any relevant news. Then generate a sample JSON payload I can use in the integration tests.”

Codex works across three tools:

Searching Google Maps for "coworking spaces Austin TX"...
Checking Google Trends for "remote work Austin"...
Searching Google News for "coworking spaces Austin"...

Results:

Maps: 8 coworking spaces found in central Austin. Top: WeWork 6th Ave (4.2★),
Capital Factory (4.6★), Industrious Domain (4.4★).

Trends: "Remote work Austin" is up 15% YoY. "Coworking Austin" up 28%.

News: Two new coworking spaces opened Q4 2025. Article from Austin American-Statesman
on downtown office vacancy driving co-working demand.

Sample JSON for integration tests:

{
  "city": "Austin",
  "results": [
    { "name": "Capital Factory", "rating": 4.6, "type": "coworking" },
    { "name": "WeWork 6th Ave", "rating": 4.2, "type": "coworking" },
    { "name": "Industrious Domain", "rating": 4.4, "type": "coworking" }
  ],
  "trend": { "direction": "up", "change": "+28%", "period": "12mo" },
  "newsCount": 2
}

Writing to tests/fixtures/local-search.json.

Codex combined three data sources and produced something directly useful. One prompt, no context-switching.

Wrapping Up

AgentPatch gives Codex access to the full tool marketplace in a single connection. Chaining tools is just what Codex does when the task calls for it — no additional configuration required. Explore everything at agentpatch.ai.

← Back to blog