@uniqent/adapter-openclaw

Install a .uniqent agent brain into OpenClaw.

Part of Uniqent — any brain, any agent: package an AI agent's whole brain into one open, signed .uniqent file and install it into whatever framework you run.

What it does

Translates a canonical .uniqent bundle into OpenClaw's native layout. The install root is the OpenClaw state dir (~/.openclaw, or $OPENCLAW_STATE_DIR):

• MCP servers → mcp.servers.<id> in openclaw.json; Telegram/Discord channels → channels.telegram / channels.discord. The file is merged, never overwritten: every other key survives, and nothing outside OpenClaw's strict config schema is written.
• credentials → ${VAR} references in openclaw.json; the values you supply at install go into the state-dir .env (merged, mode 0600) — secrets never travel in the bundle and are never inlined in the config.
• persona → SOUL.md, policies → AGENTS.md, profile → USER.md, memory facts → MEMORY.md, skills → skills/, all in the agent workspace (agents.defaults.workspace, else <state dir>/workspace). Uniqent owns only a <!-- uniqent:begin -->…<!-- uniqent:end --> section in each file and keeps your own content outside it.

Anything OpenClaw cannot represent (tasks/cron jobs, tools, runtime config, episodic memory, unmappable channels such as Slack/WhatsApp) is reported as a lossiness entry in the install plan — nothing is silently dropped.

Implements the @uniqent/adapter-sdk Adapter interface.

Install

npm install @uniqent/adapter-openclaw

Or install a brain directly via the Uniqent CLI without writing any code:

npx @uniqent/cli install my-brain.uniqent --target openclaw --root ~/.openclaw

Usage (programmatic)

import { readBundle } from '@uniqent/core';
import { openClawAdapter } from '@uniqent/adapter-openclaw';

const root = `${process.env.HOME}/.openclaw`; // the OpenClaw state dir

// 1. Load the bundle
const bundle = await readBundle('my-brain.uniqent');

// 2. Plan — inspect what will be written and what credentials are required
const plan = await openClawAdapter.plan(bundle, { root });
console.log(
  'Files to write:',
  plan.writes.map((w) => w.path),
);
console.log('Credentials needed:', plan.requiresCredentials);
console.log('Lossiness:', plan.lossiness);

// 3. Apply — merge into openclaw.json + .env and write the workspace files
//    Pass resolved credential values keyed by their ref names
const resolved: Record<string, string> = { my_api_key: 'your-key' };
const result = await openClawAdapter.apply(bundle, plan, resolved, { root });
console.log('Written:', result.written);
console.log('Notes:', result.notes);

plan() is always safe to call — it reads the bundle and your current OpenClaw files but writes nothing. apply() fails before writing anything if a required credential is missing, and is idempotent: running it twice produces byte-identical files.

Output layout

~/.openclaw/
  openclaw.json      # merged: mcp.servers.<id>, channels.<kind> (with ${VAR} refs)
  .env               # credential values, mode 0600 (merged)
  workspace/         # or agents.defaults.workspace
    SOUL.md          # persona (managed section)
    AGENTS.md        # policies (managed section)
    USER.md          # user profile (managed section)
    MEMORY.md        # memory facts (managed section)
    skills/          # one subdirectory per skill

If your openclaw.json uses JSON5-only syntax (comments, trailing commas), it is rewritten as plain JSON and the plan reports that comments are not preserved.

Exporting an existing OpenClaw setup

openClawAdapter.export({ root }) captures an existing OpenClaw setup back into a canonical Bundle — mcp.servers and Telegram/Discord channels from openclaw.json (credentials become requirements; values are never captured), plus the workspace's SOUL.md/IDENTITY.md, AGENTS.md, USER.md, MEMORY.md and skills/ — so you can re-pack and share it as a .uniqent file.

import { packBundle } from '@uniqent/core';
import { openClawAdapter } from '@uniqent/adapter-openclaw';

const bundle = await openClawAdapter.export({ root: `${process.env.HOME}/.openclaw` });
await packBundle(bundle, 'my-brain.uniqent');

License

Apache-2.0. See the Uniqent monorepo.