OpenBot

OpenBot is a local-first harness for running AI agents on your own hardware. Agents are defined with AGENT.md files and extended with plugins that handle events on the OpenBot bus. The wazoo-openbot plugin gives an OpenBot agent a persistent world-model memory backed by a hosted Wazoo world.

What the plugin provides

The plugin adds five tools to the agent that hosts it:

Tool	Purpose
`search_world`	Hybrid search over the world’s memory. Keyword search always works; the vector arm activates when embeddings are configured on the world.
`run_sparql`	Run a SPARQL 1.1 SELECT query against the world’s graph. Use this for structured questions: lists, relations, counts.
`remember`	Save one typed memory (decision, preference, person, project, finding, correction). Corrections carry the statement they supersede; statements that look like tokens or keys are refused.
`import_turtle`	Import RDF/Turtle into the world and rebuild the search index.
`reindex_world`	Rebuild the search index on demand.

Requirements

An OpenBot install (npm i -g openbot).
A Wazoo account with a platform token (wzp_...) for one-click setup, or a world token (wzw_...) and world id (w_...) to bring your own world (see tokens).

Install the plugin

Clone and build the plugin, then copy it into the OpenBot plugins folder:

git clone https://github.com/wazootech/wazoo-openbot.git
cd wazoo-openbot
bun install
bun run build
mkdir -p ~/.openbot/plugins/wazoo
cp -r dist/* ~/.openbot/plugins/wazoo/

Wire it into an agent

Reference the plugin in the agent’s AGENT.md:

plugins:
  - id: wazoo

Then connect a Wazoo platform token when the host shows the connect prompt. On first use, the plugin creates a world named after the agent, mints a scoped wzw_ token, and caches both; later calls reuse them. To pin a specific world instead:

plugins:
  - id: wazoo
    config:
      worldToken: wzw_YOUR_WORLD_TOKEN
      worldId: w_YOUR_WORLD_ID

Explicit config wins over provisioning. With neither a connection nor config, every tool call fails with a message that names both setup paths.

Memory teammate mode

The tools are passive by default: an agent uses them when it thinks to. One standing rule in each agent’s instructions turns the world into the team’s shared memory instead. Adapt this to the agent’s format:

Before work that depends on past decisions, people, preferences, or project
context, use search_world first; if the result is structural (who owns what,
list all X), use run_sparql. When something worth keeping is decided or stated,
save it with remember, filed by kind (decision, preference, person, project,
finding, correction). Never save secrets or preferences you only inferred. If a
stored fact is wrong, save a correction with supersedes set to the outdated
statement.

Every agent with the rule reads from and writes to the same world, so one agent’s decisions are visible to the rest. What the team has kept is one SPARQL query away:

PREFIX wz: <https://wazoo.dev/vocab#>
SELECT ?kind ?statement ?when WHERE {
  ?memory a ?kind ;
    wz:statement ?statement ;
    wz:recordedAt ?when .
} ORDER BY DESC(?when)

See examples/memory-teammate.md in the plugin repo for the full pattern.

Usage notes

Keyword search matches literal values in the graph, not predicate names: a search for Ethan finds triples whose object is Ethan, while a search for manages finds nothing. For structural questions (“who manages X”, “list all projects”), prefer run_sparql. import_turtle rebuilds the search index automatically. If you import through another surface, run reindex_world afterwards, or keyword search returns nothing for the new facts.

Source

The plugin lives at wazootech/wazoo-openbot. Its only runtime dependency is the Worlds REST API; tests run with bun test.

Welcome

Platform

Skills

Ecosystem

Worlds

Guides

Console

Wazoo CLI

Integrations

Comparisons

Open source

Legal

Platform API

Worlds API

What the plugin provides

Requirements

Install the plugin

Wire it into an agent

Memory teammate mode

Usage notes

Source

​What the plugin provides

​Requirements

​Install the plugin

​Wire it into an agent

​Memory teammate mode

​Usage notes

​Source

What the plugin provides

Requirements

Install the plugin

Wire it into an agent

Memory teammate mode

Usage notes

Source