Using Lindris with AI agents

The Lindris CLI is built to be driven by AI agents, not just typed by hand. Everything you can do from the terminal, an agent can do too.

Pick the page for where your agent runs:

• Claude Code — install the Lindris plugin into Claude Code.
• Codex — install it into OpenAI's Codex.
• MCP server — for Claude Desktop and any other client that speaks the Model Context Protocol, where there's no terminal to run commands in.

The first two install the same thing: an agent skill, a set of instructions the agent loads when a task matches. It teaches the agent the lindris command tree, the auth model, output flags, and ready-made recipes, so it reaches for the CLI instead of guessing at raw API calls. The agent then runs the same lindris binary you do — no extra server, no per-tool schema overhead, and it can compose full pipelines (lindris ... --output json | jq ...) the way you would.

The MCP server is a different mechanism for the case where the agent can't run shell commands at all. All three paths use the same auth and produce the same output.

What the skill covers

• Auth flow — checking lindris auth status, and asking you to run lindris auth login rather than ever handling your API key itself.
• The command tree — every live command mapped to its API endpoint, plus events simulate.
• Recipes — common pipelines like "all confirmed people on a tag, emails only" or "upsert a person with a custom field and tag".
• Guardrails — what the agent should not do (paste keys, hand-build filter[*] params, paginate manually).

Let an agent introspect the CLI

Any agent that can run the CLI can discover its full surface without parsing --help, with agent-context:

lindris agent-context --pretty

This emits a machine-readable JSON description of every command, flag, and the auth model, versioned by schema_version — a stable contract an agent can read at runtime.

Another agent entirely

The skill uses the portable Agent Skills SKILL.md format, so an agent we don't special-case can still use it. Copy the skill directory out of the published package into whatever path that agent reads:

npm pack @lindris/claude-plugin --pack-destination /tmp
tar -xzf /tmp/lindris-claude-plugin-*.tgz -C /tmp
cp -R /tmp/package/skills/using-lindris-cli <that agent's skills directory>

Run lindris agents list to see which hosts have first-class support.

Where to next

• Claude Code — plugin install for Claude Code.
• Codex — plugin install and the bare-skill fallback.
• MCP server — lindris-mcp for Claude Desktop and other MCP clients.
• Authentication — the key-resolution model every path shares with the CLI.