Lerryn
简体中文

A Niltavi product

A steady place
for agent work.

Your agent starts the work. Lerryn gives it a sandbox to run in, credentials by reference, and a clear record of what happened.

Start self-hosting Explore the source
  • Python · Node · Shell
  • MCP · HTTP API · CLI
  • Self-hosted
Your agent

Prepare the daily report.

Lerryn sandbox Completed

The daily report

daily-report · Python

  1. 01Fetch records
  2. 02Build report
  3. 03Save artifact
Back to your agent

report.csv + execution log

From request to result

Keep the work flowing.
Keep the details close.

Connect your agent

Use MCP, the HTTP API, or the CLI to apply a task and start a run. Add a schedule or webhook when the work needs to repeat.

Run on your infrastructure

Execute Python, Node, and Shell in Docker containers. Start with SQLite on one host, then use PostgreSQL and S3 for a multi-node deployment.

Follow every run

See stages, logs, results, and artifacts in the web console. Keep execution history with the task, so your next decision has context.

One platform, familiar interfaces

Meet your agent
where it works.

Use a local MCP connection, inspect the HTTP API, or work directly from your terminal. Each route reaches the same task and run model.

Read the CLI & agent guide

Connect Claude Code through the CLI

lerryn login https://your-lerryn.example --username alice
claude mcp add lerryn -- lerryn mcp --stdio

Remote MCP clients can also connect to your deployment’s /mcp endpoint.

Discover what your deployment supports

curl https://your-lerryn.example/api/v1/discover \
  -H "Authorization: Bearer $LERRYN_TOKEN"

The OpenAPI contract describes the operations, schemas, and errors.

Apply a task, then run it

lerryn tasks apply ./daily-report
lerryn runs start daily-report --test --wait 60s

Keep the task manifest and files together in a versioned directory.

Your infrastructure. Your controls.

Give tasks access.
Keep credentials in place.

Bind credentials by reference and inject them at runtime. Lerryn encrypts stored values and redacts known secret values from captured logs. Task code still needs the same care and review as any code with access to your systems.

Credential references
Agents request access by name.
Execution records
Stages, results, logs, and artifacts.
Self-hosted storage
SQLite or PostgreSQL. Local files or S3.

Give your next task
a place to run.

Start with a single-node deployment. Connect an agent when you’re ready.

Start self-hosting