Kode
A terminal coding agent with native AGENTS.md support, subagents, and 20+ model providers.
About
Kode is an open-source terminal coding agent from ShareAI Lab that leans hard into the AGENTS.md standard — the open format for guiding coding agents that 60k+ repos have adopted. It reads project instructions walking up from the git root, supports AGENTS.override.md, stays compatible with legacy .claude directories and CLAUDE.md, and can generate and maintain your AGENTS.md for you.
Beyond the standards story, it is a full agent: subagent delegation and task orchestration, 20+ model providers, todo planning, plan mode, custom commands, and memory compression. Distribution is thoughtful — per-platform native binaries via npm optionalDependencies with a Node.js fallback, plus standalone GitHub release binaries.
The honest caveat: "design for post-human workflows" is the kind of tagline that should make you reach for specifics, and the specifics are solid but familiar — this is a well-built entry in a crowded CLI-agent field, differentiated mainly by its AGENTS.md depth. If your repos already standardize on AGENTS.md, that depth is exactly the reason to pick it.
What it can do
- Native AGENTS.md standard support (plus .claude/CLAUDE.md compatibility)
- Subagent system for delegation and task orchestration
- 20+ AI model providers
- Todo planning, plan mode, custom commands, memory compression
- Non-interactive run mode for scripting
- Per-platform native binaries; Apache-2.0 licensed
Setup
- Install:
npm install -g @shareai-lab/kode(per-platform native binary, Node.js fallback), Homebrew, or a standalone binary from GitHub Releases. - Configure credentials for your model provider(s) — 20+ supported.
- Run in your repo; Kode picks up AGENTS.md / AGENTS.override.md automatically.
- Use
#requests to generate docs, subagents for delegation, and plan mode for reviewed execution.
Prerequisites
Node.js (for the npm path) or a supported OS for native binaries; API credentials for at least one model provider.
Use cases
Works with
Trust score — 8/10
Scored 8/10 (metrics tier, metrics-v1). Evidence: Adoption 7: starScore from 5,235 stars on shareAI-lab/Kode-cli (log scale vs 298k peer max); Maintenance 9: pushed 2026-10-09 (recency 10); active releases incl. native binary distribution; Openness 10: Apache-2.0 (0.7*10) + official docs verified (0.3*10); Manual 7: AGENTS.md-native design corroborated by README and agents.md standard; docked for marketing fluff ('post-human workflows') and crowded field. License: Apache-2.0.
★ 5,235 GitHub stars
How the trust score is built
Full methodology →- Adoption · stars/downloads, log-scaled vs peers
- 7.0/10
- Maintenance · recency, releases, issue load
- 9.0/10
- Openness · license + docs
- 10.0/10
- Manual review · sanity read + red-flag check
- 7.0/10
- · Adoption 7: starScore from 5,235 stars on shareAI-lab/Kode-cli (log scale vs 298k peer max)
- · Maintenance 9: pushed 2026-10-09 (recency 10); active releases incl. native binary distribution
- · Openness 10: Apache-2.0 (0.7*10) + official docs verified (0.3*10)
- · Manual 7: AGENTS.md-native design corroborated by README and agents.md standard; docked for marketing fluff ('post-human workflows') and crowded field
Sources
Every fact above was checked against these sources. Where a claim comes from the provider and couldn't be independently verified, the text says so.