Livebook

Interactive web-based notebook for Elixir — the Jupyter equivalent for the Elixir/BEAM ecosystem. Livebook runs code cells in a persistent Elixir runtime (full BEAM VM), supports real-time collaborative editing, and integrates with Nx (numerical computing), Bumblebee (HuggingFace ML models), Kino (UI widgets, charts, dataframes), and VegaLite for visualization. Smart Cells provide no-code UI for common operations (database queries, API calls, ML model inference). Deployable as standalone app (Livebook Desktop), Docker container, or Fly.io deployment. Used for ML exploration, API prototyping, team documentation, and data analysis.

Evaluated Mar 06, 2026 (0d ago) v0.14.x
Homepage ↗ Repo ↗ Developer Tools elixir notebook interactive ml data-science collaboration livebook beam smart-cells
⚙ Agent Friendliness
66
/ 100
Can an agent use this?
🔒 Security
83
/ 100
Is it safe for agents?
⚡ Reliability
80
/ 100
Does it work consistently?

Score Breakdown

⚙ Agent Friendliness

MCP Quality
--
Documentation
88
Error Messages
90
Auth Simplicity
85
Rate Limits
92

🔒 Security

TLS Enforcement
90
Auth Strength
80
Scope Granularity
75
Dep. Hygiene
88
Secret Handling
82

Access token auth for local use. Livebook Teams adds SSO. Code cells run with full BEAM process permissions — trust all notebook authors. Secrets stored in Livebook secret store, not in notebook files.

⚡ Reliability

Uptime/SLA
82
Version Stability
80
Breaking Changes
75
Error Recovery
85
AF Security Reliability

Best When

You're an Elixir developer wanting interactive exploration of ML/data workflows, team-executable runbooks, or a scratchpad for BEAM-powered computation without leaving the Elixir ecosystem.

Avoid When

You're doing Python-centric data science (use Jupyter), need production deployment, or your team has no Elixir knowledge. Livebook complements Elixir development; it doesn't replace Python data science tooling.

Use Cases

  • Prototype agent LLM workflows interactively in Elixir using Livebook + Bumblebee — run HuggingFace model inference, explore embeddings, and test prompts without setting up a full application
  • Build interactive agent dashboards using Livebook's Kino widgets — real-time data tables, charts, and form inputs for monitoring agent system state
  • Explore and document Nx numerical computing operations for agent data preprocessing — tensor manipulations, statistical operations, and vectorized transformations in interactive cells
  • Create team runbooks and automated workflows in Livebook — executable documentation where teammates can run agent operations with one click using Smart Cells
  • Use Livebook as a REPL/scratchpad for Elixir Phoenix/Ecto — connect to development database with Smart Cell, run Ecto queries, inspect LiveView state

Not For

  • Production application deployment — Livebook is an exploration and prototyping tool; production code should be in Mix projects; Livebook scripts are not meant for production serving
  • Non-Elixir data science — Jupyter with Python (pandas, scikit-learn) has dramatically larger ecosystem for data science; Livebook is for Elixir teams extending into ML, not Python data scientists switching
  • Teams without Elixir knowledge — Livebook requires Elixir/BEAM knowledge; unlike Jupyter where Python familiarity is sufficient, Livebook's value compounds with Elixir expertise

Interface

REST API
Yes
GraphQL
No
gRPC
No
MCP Server
No
SDK
Yes
Webhooks
No

Authentication

Methods: password
OAuth: No Scopes: No

Livebook uses a password or token for local and hosted access. Livebook Teams (paid) adds team auth via OAuth/SSO. Default local mode uses a single access token in the URL.

Pricing

Model: open_source
Free tier: Yes
Requires CC: No

Livebook open source is free. Livebook Teams adds collaboration, deployment, and enterprise features. Fly.io integration available for cloud deployment with Fly's pricing.

Agent Metadata

Pagination
none
Idempotent
Partial
Retry Guidance
Not documented

Known Gotchas

  • Cell execution order matters — Livebook cells share a single BEAM process; variables defined in earlier cells are visible to later cells only if cells are evaluated in order; 'Evaluate all' runs in document order
  • Runtime state is lost on disconnect — unlike Jupyter which can persist kernel state between browser sessions, Livebook runtime state resets on disconnect or timeout; save important results to files or databases
  • Bumblebee model downloads on first use — Bumblebee downloads HuggingFace models to local cache on first Smart Cell use; large models (GPT-2, BERT) can be 500MB+; ensure disk space and network access
  • EXLA compilation on first use — Nx with EXLA backend JIT-compiles operations on first run; initial tensor operations may take 30-60 seconds to compile; subsequent runs are fast
  • Collaborative editing via Fly.io requires configuration — Livebook Teams or manual Fly.io setup needed for shared notebooks; local Livebook is single-user by default; access token in URL is the only auth
  • Smart Cells require specific dependencies — database Smart Cells need kino_db in Mix.install; ML Smart Cells need bumblebee + axon; missing dependencies cause Smart Cell to fail without clear error

Alternatives

Full Evaluation Report

Detailed scoring breakdown, competitive positioning, security analysis, and improvement recommendations for Livebook.

$99

Scores are editorial opinions as of 2026-03-06.

5215
Packages Evaluated
26151
Need Evaluation
173
Need Re-evaluation
Community Powered