
Mercury Agent adds Copilot, Codex support
Mercury Agent 1.1.7 adds GitHub Copilot and OpenAI Codex as first-class providers, letting one workflow tap multiple model ecosystems without changing tools. The release also improves live progress tracking, loop detection, and Telegram handling for more controlled long-running agent runs.
Mercury is pushing past "single-provider agent" territory and into orchestration territory, which is the more defensible long-term wedge for this category.
- –Copilot and Codex support makes Mercury more attractive to teams already paying for those ecosystems, especially if they want one operator loop across models
- –The guardrails matter as much as the integrations: permission approvals, token controls, and background execution are the difference between a demo and a daily driver
- –Real-time progress and loop detection are the right kind of plumbing improvements for agentic work, where failures are usually about state, not raw capability
- –If Mercury can keep provider abstraction clean, it can become a control layer; if not, it risks becoming another brittle wrapper around vendor APIs
- –This is a meaningful update for AI coding workflows, but it is still an incremental product update rather than a new category breakout
DISCOVERED
1h ago
2026-05-10
PUBLISHED
2h ago
2026-05-10
RELEVANCE
AUTHOR
mercury__agent