kelex-cli 0.1.0

Terminal interface to Kelex — chat with your Lara persona, run swarms, route to frontier brains, all from the shell.

kelex-cli

Terminal interface to Kelex — your AI agent with persistent memory, parallel workers, and real tools.

$ pip install kelex-cli
$ kelex login
  paste your tenant ID + api token (one-time)
$ cd my-project
$ kelex
  Lara · this-project · 0 turns
  > what does foo.py do?
  ...

Why use this over claude / aider / etc.

• Persistent memory across sessions — Lara remembers your project from last week without you re-pasting context.
• Native swarm — > spawn 5 workers to review this PR runs 5 parallel reasoners, then synthesizes one answer.
• Multi-brain routing — > ask kimi about this segfault routes to Kimi K2.6 mid-conversation, without leaving Lara's loop.
• Self-scheduling — > remind me every Monday 9am to triage issues sets a real cron that wakes Lara up in the cloud.
• Self-hostable — point kelex at your own Kelex instance, not someone else's API.

Quick start

pip install kelex-cli
kelex login --base https://kelex.brownforces.io
# paste your tenant_id and api_token (one-time, stored at ~/.kelex/config.toml)
cd ~/code/my-project
kelex

Project-scoped agents

The first time you run kelex in a directory, it creates a project-scoped agent named after the directory. Each project gets its own Lara who remembers that project specifically. Switch projects = switch context.

cd ~/code/api      # uses agent "api-lara" (auto-created)
cd ~/code/frontend # uses agent "frontend-lara" (auto-created)

Or pin an existing agent:

kelex --agent <agent_id>

File attach

Inline @path/to/file.ext in any prompt to send the file's contents to Lara:

> compare @src/auth.py with @src/auth_v2.py and tell me what changed

Env vars

KELEX_BASE       # default: https://kelex.brownforces.io
KELEX_TENANT_ID  # your tenant UUID
KELEX_API_TOKEN  # your api token
KELEX_AGENT_ID   # pin a specific persona (otherwise auto-resolved per cwd)

License

MIT.