A small set of agent-agnostic skills that help AI coding assistants (Claude Code, Cursor, Codex, Continue, Aider, etc.) understand and work with Unkey — both API Management (key issuance, rate limiting, identities) and Unkey Deploy (container hosting for your backend).
Each skill is a single Markdown file with optional YAML frontmatter. Any agent that can read instructions from a file (or fetch from a URL) can use them.
| Skill | What it does |
|---|---|
unkey-overview/SKILL.md |
Explains what Unkey is, the difference between API Management and Deploy, current pricing, and where the canonical docs live. Use when a user asks about Unkey itself. |
unkey-api-management/SKILL.md |
Guides an agent through integrating Unkey API Management — issuing/verifying API keys, standalone rate limiting, identities, RBAC (roles & permissions), analytics, and audit logs. Use when a user wants to add auth, rate limits, or per-customer quotas in front of an API they already run. |
unkey-deploy/SKILL.md |
Guides an agent through deploying a user's app to Unkey Deploy — detects existing build config (Dockerfile, Railpack, buildpack), converts or generates a Dockerfile when needed, handles CLI install and auth, and runs the deploy. |
Clone into your skills directory (user-level or project-level):
# User-level — available in every project
git clone https://github.com/unkeyed/skills ~/.claude/skills/unkeyed
# Or project-level — committed alongside your codebase
git clone https://github.com/unkeyed/skills .claude/skills/unkeyedClaude Code will pick up SKILL.md files inside any subdirectory. Restart
the session (or /reload) and the skills appear in the skill list.
Cursor reads project rules from .cursor/rules/*.mdc. Copy the body of each
SKILL.md (or the whole file, renamed) into that directory:
mkdir -p .cursor/rules
cp unkey-overview/SKILL.md .cursor/rules/unkey-overview.mdc
cp unkey-deploy/SKILL.md .cursor/rules/unkey-deploy.mdcMost agents accept a system prompt, custom instructions file, or @file
reference. Either:
- Point your agent at the raw GitHub URL of the skill you want, e.g.
https://raw.githubusercontent.com/unkeyed/skills/main/unkey-deploy/SKILL.md - Or paste the contents into your agent's custom instructions or rules
file (e.g.
AGENTS.md,.aider.conf.yml'sread:,CONTINUE.md).
For Codex CLI specifically, drop the file into your project's AGENTS.md or
reference it from there.
Each SKILL.md is a self-contained prompt. Feed it to any model alongside
the user's request and the model will follow it. The YAML frontmatter is
informational — agents that don't recognize it will safely ignore it.
The three skills are designed to hand off to each other:
unkey-overviewis informational. If a user asks "what is Unkey?", "how much does it cost?", or "should I use Unkey for X?", this is the right skill. It explicitly does not write code or run commands. It routes to the other two skills once the user has picked a product.unkey-api-managementis action-oriented. It assumes the user has decided to add API keys, rate limiting, identities, or RBAC in front of an API they already run. It walks them from "no SDK installed" to "verifying real requests against Unkey."unkey-deployis action-oriented. It assumes the user has decided to host their service on Unkey. It walks them from "no Dockerfile" to "running container" with the minimum number of questions.
If you're authoring an agent integration, you can load all three — they reference each other by file path so the agent knows when to switch.
These skills will drift as Unkey's pricing, CLI, and Dockerfile guidance evolve. If you spot something stale, open an issue or PR.
Canonical sources the skills point at (in case they move):
https://unkey.com/llms.txt— agent-oriented docs indexhttps://unkey.com/pricing.md— current pricinghttps://www.unkey.com/docs/builds/dockerfile-prompt.md— Dockerfile generation guidehttps://www.unkey.com/docs— human docs
MIT.