Data enrichment for Codex
Run waterfall enrichment from Codex. Same tables, same providers, same governed spend, driven from OpenAI's coding agent.
you › Add a verified email and company headcount to every row in my "Inbound signups" table, cap 2,000 credits.
oxygen › Codex calls Oxygen's enrichment tools; Oxygen waterfalls each row, stops at the cap, and records source and cost per cell for you to approve.
Oxygen is agent-client agnostic: if your team lives in Codex rather than Claude Code, the enrichment workflow is identical. Codex calls the same Oxygen tools over MCP, on the same tables.
Describe the columns you want in Codex, and Oxygen waterfalls across 30+ providers, applies your spend cap, and records the source and cost on every cell, on a Postgres database your organization owns.
How it works
- 1
Connect Oxygen to Codex
Add the Oxygen MCP server to Codex. The tool catalog, tables, and auth are the same as any other surface.
- 2
Ask for the columns
Describe the fields; Oxygen maps them to enrichment columns and chooses the provider waterfall.
- 3
Run under a cap
Preview the cost, then run with a spend cap. Providers fall back automatically.
- 4
Approve and sync
Review provenance in the web app, then export or sync onward.
What you get
- Identical enrichment behavior across Codex, Claude Code, and the CLI
- Waterfall across 30+ providers with automatic fallback
- Spend caps, dry-run previews, and per-cell provenance
- Bring your own keys or use managed credits
- Data stays in a Postgres database you own
Frequently asked questions
Does Oxygen work the same in Codex as in Claude Code?
Yes. Oxygen is MCP-first and client-agnostic, so the tools, tables, auth, and governed spend are identical whether you drive it from Codex, Claude Code, or the CLI.
Do I need to write code to enrich in Codex?
No. You describe the columns in plain language and Codex calls Oxygen's enrichment tools. Code-authored workflows are available if you want them, via durable recipes.
Where does the enriched data live?
In your organization's own Postgres database, with the source and cost recorded on every cell.
Run data enrichment from your agent
Point Codex at Oxygen and let it do the work, on a database you own.