Codex Agent Harness | HarnessRouter
Browse 41 harnesses
30-day usage· Aug 12–Sep 10, 2026
Data as of September 11, 2026 · About the data
Rank #9#8 of21CLI agents · 7-day #8
Tokens 3.6T
Requests 50.4M
Capabilities and verification
Reviewed against the HarnessRouter capability catalog on August 24, 2026. Every verified row cites its public source; rule-derived rows apply HarnessRouter's published classification rules to documented behavior. A “No public evidence” badge marks a gap in the public record, not a statement that Codex lacks them.
Configuration & support
- Other models Directly verified
- Custom instructions Directly verified
- Skills Directly verified
- Custom tools No public evidence
- MCP Directly verified
- Sandbox & approvals Directly verified
- Sessions Directly verified
- Subagents Directly verified
Execution & agency
- Workspace files Directly verified
- Structured actions Directly verified
Environment control
- Browser control Directly verified
- Computer control Directly verified
Observation & capture
- Screen observation Directly verified
- Screenshot export Directly verified
- Recording export No public evidence
Artifact outputs
- Documents Directly verified
- Spreadsheets Directly verified
- Presentations Directly verified
- Images Directly verified
- Video Rule-derived
- App deployment Directly verified
Sources: openai.com · developers.openai.com
Machine interfaces
- Headless CLI · Stable
Benchmark evidence
Codex configurations from the HarnessRouter Care Prep benchmark, published August 10, 2026. One controlled synthetic-data task; results apply to this test, not universally.
| Model | End-to-end latency | Cost |
|---|---|---|
| gpt-5.2 | 4m 36s | 0.72 credits |
| gpt-5.5 | 2m 46s | 59.7 credits |
Execution success and strict grounding pass rates for these configurations are published in the observed-history capture on the benchmark report.
Run Codex through HarnessRouter
API harness ID: codex
Use it for code-related work that your product can review.