Loom tutorials

These prompt-first tutorials use the small public Python palmerpenguins repository because it has real package code, tests, and bundled CSV data without requiring a private repository.

You don't need to memorize commands. Practice a repeatable conversation in your coding harness: ask a normal coding question, let the agent check Loom, and confirm that its answer reflects current shared memory. The prompts work in both Claude Code and Codex unless a tutorial calls out a harness-specific step.

Everything you build here runs on the same memory model behind Loom's benchmark scores on STALE, LongMemEval, and AMA-Bench.

How the path works

• Start with Tutorial 00 to connect Claude Code or Codex and verify Loom. It points you to the signed-in install page and separates the few installation and hook steps that differ by harness.
• Learn how named stores separate your normal project memory, safe tutorial data in [loom_demo], and multi-agent coordination in [kanban].
• Use the public test repository and writable test store to discover Entity and Relation types, then work through real coding-memory examples.
• Build from current Decisions and Roadmap items, calculate with typed historical data, promote repeated queries into rules, and finally coordinate an agent fleet.

The order is deliberate. First you establish a safe connection and learn how Loom organizes memory. Then you practice saving and retrieving project knowledge, letting relevant context resurface automatically, and turning repeated work into reusable capabilities. The final tutorials apply those skills to planning, maintenance, and multi-agent coordination.

Store names vary by account. Tutorial 02 shows how to list, load, describe, and unload your stores without replacing the default store for the session.

Tutorial format

Every walkthrough uses the same three-part format:

End every session with a capture pass

Memory compounds when you close the loop. At the end of a working session, use this prompt:

Run this as a habit, not as a demand to create a fact every time.

Follow the tutorial path

• Tutorial 00: Connect your coding harness to Loom

  Connect your coding harness, enable Loom's hooks, and verify that shared memory is ready.

• Tutorial 01: Build your first memory in five minutes

  Save useful project context and see it return in a fresh coding session.

• Tutorial 02: Connect to multiple stores

  Connect one agent to multiple stores while keeping each memory boundary clear.

• Tutorial 03: Set up project memory

  Give Loom enough project context to return the right memory at the right time.

• Tutorial 04: Understand entities, relations, and types

  Learn how entities, relations, and built-in types turn notes into connected knowledge.

• Tutorial 05: Capture directives and findings

  Capture standing instructions and verified discoveries so agents stop relearning them.

• Tutorial 06: Use proactive memory

  See how Loom surfaces relevant memory during work and carries unfinished work forward.

• Tutorial 07: Create a custom type

  Create a reusable type when your team needs to remember domain-specific information.

• Tutorial 08: Plan with past decisions

  Build a coding plan from current roadmap items and decisions instead of starting cold.

• Tutorial 09: Promote repeated work into a rule

  Turn repeated reasoning into a reusable rule that saves time and tokens.

• Tutorial 10: Coordinate multiple agents

  Coordinate several agent sessions through a shared Loom Kanban board.

• Tutorial 11: Maintain Loom memory

  Keep shared memory accurate, preserve completed work, and record delivery evidence.

• Tutorial 12: Audit and maintain your Loom store

  Run a repeatable store-health audit, apply bounded repairs, and verify the result without erasing history.

Back to Loom Cloud