Breaking

Collection of Over 166 Specialized Sub-Agents for Codex Released

July 2, 2026 at 19:31 EDT

AI Agents
Software Dev & Coding
Open Source

A free, community-maintained collection of 166–171+ specialized subagents for OpenAI Codex, called "awesome-codex-subagents," has been published on GitHub under the stewardship of VoltAgent. Spanning 13 development categories, it distributes each agent as a .toml custom-definition file.

OpenAI Codex · Subagents

166+ Specialized Subagents, Bundled for Codex in One Drop-In Library

A new open collection packages ready-to-use subagent definitions across 13 development categories — copy the .toml files into your Codex setup and delegate work to parallel, role-tuned agents.

171+

Subagent definitions (166 at launch)

Development categories

Models routed by purpose

.toml

Definition file format

How work flows through the main agent

Enabled by default — one agent spins up specialists in parallel, then merges results.

Main Agent

Receives the task

→

Subagents

Run in parallel by role

→

Merge

Combined response

Models routed by purpose

Each definition can assign a cheaper, faster model to lightweight roles — reserving deep reasoning for where it counts.

gpt-5.4

Deep reasoning

gpt-5.3-codex-spark

Fast & lightweight

Taller column = heavier reasoning depth per task.

Praised

Assign cheaper models by role
Useful on large codebases
Easy custom definitions via TOML
Moves beyond single-agent dependence

Limitations noted

Hard to launch without naming the target
Token consumption rises
Confusion when files overlap
Unsuited to small tasks; IDE visibility "coming soon"

Install

Copy .toml files into ~/.codex/agents/ (global) or .codex/agents/ (per project). Each defines role instructions, a checklist, model, reasoning depth, and a sandbox_mode — read-only or workspace-write.

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.