Skip to content

Open-source platform for 3D AI agents. Turn text or a photo into a rigged, animated GLB avatar, give it an LLM brain, memory and a wallet, and embed it anywhere with one web component. Remote MCP server, x402 payments, Three.js, browser-native.

docs(packages): state the proprietary license and document the voice client, probes, and example

nirholasc337a34a04bf
diff --git a/packages/naming-mcp/README.md b/packages/naming-mcp/README.md
index 33e28cbb2..8f4e7723c 100644
--- a/packages/naming-mcp/README.md
+++ b/packages/naming-mcp/README.md
@@ -121,7 +121,7 @@ All three are live public reads against Solana mainnet. A `.sol` name with no ow
- Homepage: https://three.ws
- Changelog: https://three.ws/changelog
- Issues: https://github.com/nirholas/three.ws/issues
-- License: Apache-2.0 — see [LICENSE](./LICENSE)
+- License: proprietary, all rights reserved. See [LICENSE](./LICENSE).
---
diff --git a/packages/portfolio-mcp/README.md b/packages/portfolio-mcp/README.md
index b3673422a..74e28e0ca 100644
--- a/packages/portfolio-mcp/README.md
+++ b/packages/portfolio-mcp/README.md
@@ -135,7 +135,7 @@ For an SPL transfer it reads the mint's decimals on-chain, verifies your balance
- Homepage: https://three.ws
- Changelog: https://three.ws/changelog
- Issues: https://github.com/nirholas/three.ws/issues
-- License: Apache-2.0 — see [LICENSE](./LICENSE)
+- License: proprietary, all rights reserved. See [LICENSE](./LICENSE).
---
diff --git a/packages/pumpfun-mcp/README.md b/packages/pumpfun-mcp/README.md
index 636acd8b7..d61217f61 100644
--- a/packages/pumpfun-mcp/README.md
+++ b/packages/pumpfun-mcp/README.md
@@ -169,7 +169,7 @@ Failures surface as MCP tool errors (`isError: true`) with the real cause — ne
- Homepage: https://three.ws
- Changelog: https://three.ws/changelog
- Issues: https://github.com/nirholas/three.ws/issues
-- License: Apache-2.0 — see [LICENSE](./LICENSE)
+- License: proprietary, all rights reserved. See [LICENSE](./LICENSE).
---
diff --git a/packages/vision-mcp/package.json b/packages/vision-mcp/package.json
index f1512286a..4844f8858 100644
--- a/packages/vision-mcp/package.json
+++ b/packages/vision-mcp/package.json
@@ -30,6 +30,7 @@
},
"files": [
"src",
+ "examples",
"server.json",
"README.md",
"LICENSE"
diff --git a/packages/voice/README.md b/packages/voice/README.md
index 9aec774aa..a0687eec4 100644
--- a/packages/voice/README.md
+++ b/packages/voice/README.md
@@ -187,13 +187,37 @@ than picking your own TTS — one round trip instead of two.
Fetch the live voice catalog (`GET /api/tts/voices`) — ids, names, descriptions,
which synthesis lanes are configured. Render a picker before the user commits.
-### Capability probes
+### Capability probes: `asrInfo()` / `lipsyncInfo()`
Each lane answers a `GET` probe so a UI can decide whether to use the server lane
or its in-browser fallback without sniffing the browser:
-- `GET /api/asr` → `{ configured, encodings, sampleRate }`
-- `GET /api/a2f` → `{ configured, canSynthesize, model, fps, blendshapeFormat: 'arkit', sampleRate, accepts }`
+```js
+import { asrInfo, lipsyncInfo } from '@three-ws/voice';
+
+const asr = await asrInfo(); // GET /api/asr
+// → { configured, encodings: ['wav','pcm','flac','ogg'], sampleRate, raw }
+
+const face = await lipsyncInfo(); // GET /api/a2f
+// → { configured, canSynthesize, model, fps, blendshapeFormat: 'arkit', sampleRate, accepts, raw }
+```
+
+`configured: false` means that lane's provider key isn't set on the deployment,
+so the call would raise `not_configured` — render the in-browser fallback
+instead. `raw` carries the untouched endpoint response for anything not shaped
+above.
+
+### `createVoice(options?) → client`
+
+Bind every call above to one configuration (a custom `baseUrl`, a bearer token, a
+payment-aware `fetch` for the paid backstops) and reuse it:
+
+```js
+import { createVoice } from '@three-ws/voice';
+
+const voice = createVoice({ baseUrl: 'https://three.ws', apiKey: process.env.THREE_WS_API_KEY });
+const clip = await voice.speak('Bound to one config.');
+```
## Voices
@@ -280,6 +304,13 @@ you pass pre-synthesized audio to `lipsync()` instead.
## Examples
+A runnable script that exercises the whole loop against the live endpoints lives
+in [`examples/`](./examples):
+
+```bash
+node examples/voice-loop.mjs # probe → speak → lipsync → transcribe → say
+```
+
**Browser voice loop → animated avatar.** Record, transcribe, answer, and drive
the face on a loaded GLB:
diff --git a/packages/voice/package.json b/packages/voice/package.json
index efcab8eb0..62afe5e09 100644
--- a/packages/voice/package.json
+++ b/packages/voice/package.json
@@ -16,6 +16,7 @@
},
"files": [
"src",
+ "examples",
"README.md",
"LICENSE"
],
diff --git a/packages/x402-mcp/package.json b/packages/x402-mcp/package.json
index 0b4e7082c..0a34cc013 100644
--- a/packages/x402-mcp/package.json
+++ b/packages/x402-mcp/package.json
@@ -33,6 +33,7 @@
},
"files": [
"src",
+ "examples",
"server.json",
"README.md",
"LICENSE"