Skip to content

Commit 3293370

Browse files
cjagwaniclaude
andcommitted
docs(platform): publish canonical platform support matrix (#4630)
Adds docs/reference/platform-support.mdx as the single source of truth for what NemoClaw supports today across platforms, inference providers, agents, messaging integrations, deployment paths, capabilities, and out-of-scope items. Extends ci/platform-matrix.json with the new sections, expands the generator to render them, and wires the new file into the pre-commit sync hook and the docs nav (both agent variants). Status vocabulary now covers all five categories from the issue spec: Tested, Tested with limitations, Experimental, Deferred, Unsupported. The platforms table gains a CI column so 'Tested with limitations + not in CI' (Windows WSL2) reads precisely. Local inference rows carry validated container image pins, model defaults per host, GPU memory minimums for NIM, NVIDIA Container Toolkit / CDI runtime assumption, and NGC registry authentication notes. Capabilities cover the eight launch-facing claims surfaced in overview.mdx: guided onboarding, sandboxed execution, routed inference, declarative network policy, snapshot and restore, agent skills, state migration, blueprint versioning, plus web search backend. Hermes is demoted to 'Tested with limitations' with citations to active deployment-identity bugs (#5211, #5327) and the empty model-provider compatibility registry; OpenClaw stays Tested. Out of scope enumerates 12 items with file:line citations so support triage can distinguish bugs from unsupported-config requests. Cross-linked from README, docs/index.mdx, docs/about/overview.mdx, docs/about/release-notes.mdx, and docs/get-started/prerequisites.mdx. Pre-commit hook (.pre-commit-config.yaml) extended so edits to the JSON or the new page trigger regeneration. CI runs the generator's --check mode in .github/workflows/pr.yaml so the rendered tables never drift from the JSON silently. Open items intentionally left for PR review: - Product owner is recorded as 'TBD (see PR review)'. The engineering owner team is named via CODEOWNERS; the product owner needs to be filled in by the launch / PM lead before this page is referenced from launch-facing material. - Alignment with any external PRD, launch deck, or sales material is out of scope for the doc itself. PR reviewers from product / launch should confirm the matrix does not conflict with public claims being prepared elsewhere. Refs #4630. Signed-off-by: Charan Jagwani <[email protected]> Co-Authored-By: Claude Opus 4.7 <[email protected]>
1 parent 7104567 commit 3293370

11 files changed

Lines changed: 596 additions & 17 deletions

File tree

‎.pre-commit-config.yaml‎

Lines changed: 2 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -81,9 +81,9 @@ repos:
8181
hooks:
8282
- id: platform-matrix-sync
8383
name: Sync platform matrix to docs
84-
entry: bash -c 'python3 scripts/generate-platform-docs.py && git add docs/get-started/prerequisites.mdx docs/inference/inference-options.mdx'
84+
entry: bash -c 'python3 scripts/generate-platform-docs.py && git add docs/get-started/prerequisites.mdx docs/inference/inference-options.mdx docs/reference/platform-support.mdx'
8585
language: system
86-
files: ^(ci/platform-matrix\.json|docs/get-started/prerequisites\.mdx|docs/inference/inference-options\.mdx|scripts/generate-platform-docs\.py)$
86+
files: ^(ci/platform-matrix\.json|docs/get-started/prerequisites\.mdx|docs/inference/inference-options\.mdx|docs/reference/platform-support\.mdx|scripts/generate-platform-docs\.py)$
8787
pass_filenames: false
8888
priority: 3
8989

‎README.md‎

Lines changed: 1 addition & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -59,6 +59,7 @@ NemoClaw is an alpha project, so maintainers review issues, discussions, and pul
5959
| Reproducible bugs | [GitHub Issues](https://github.com/NVIDIA/NemoClaw/issues) |
6060
| Feature proposals | Start with [GitHub Discussions](https://github.com/NVIDIA/NemoClaw/discussions), then open an issue when the scope is clear |
6161
| Current priorities | [Current Priorities](#current-priorities) |
62+
| Supported platforms, providers, and integrations | [Platform Support and Launch Claims](docs/reference/platform-support.mdx) |
6263
| Contribution help | [CONTRIBUTING.md](CONTRIBUTING.md) |
6364
| Security vulnerabilities | Use the private channels in [SECURITY.md](SECURITY.md); do not open public issues |
6465

‎ci/platform-matrix.json‎

Lines changed: 196 additions & 9 deletions
Original file line numberDiff line numberDiff line change
@@ -1,13 +1,28 @@
11
{
2-
"$comment": "Single source of truth for tested platforms and providers. Scripts read this to generate README and docs tables. QA/CI update this file; docs are derived.",
3-
"version": "1.0",
4-
"updated": "2026-04-08",
2+
"$comment": "Single source of truth for NemoClaw launch claims and platform support. Covers platforms, inference providers, supported agents, messaging integrations, and deployment paths. Scripts read this to generate README and docs tables. QA/CI update platform/provider rows; the engineering owner reviews other rows. Docs are derived.",
3+
"version": "1.1",
4+
"updated": "2026-06-12",
5+
6+
"project_status": {
7+
"stage": "alpha",
8+
"label": "Early preview",
9+
"since": "2026-03-16",
10+
"notes": "Maintainers review issues, discussions, and PRs on a best-effort basis without guaranteed response timelines."
11+
},
12+
13+
"owners": {
14+
"engineering": "@NVIDIA/nemoclaw-maintainer",
15+
"product": "TBD (see PR review)",
16+
"$comment": "Engineering owner is the GitHub team that auto-reviews changes to this file via CODEOWNERS. Product owner is the human who signs off on launch-facing claim changes before they reach demos or sales material."
17+
},
518

619
"statuses": {
720
"tested": "Validated by CI/QA. Primary path.",
821
"caveated": "Tested with limitations. See Notes.",
922
"experimental": "Requires NEMOCLAW_EXPERIMENTAL=1.",
10-
"deferred": "Planned but not yet validated."
23+
"deferred": "Planned but not yet validated.",
24+
"unsupported": "Explicitly out of scope. Not validated and not planned.",
25+
"hermes only": "Available only when onboarding the Hermes agent."
1126
},
1227

1328
"platforms": [
@@ -17,7 +32,7 @@
1732
"status": "tested",
1833
"prd_priority": "P0",
1934
"ci_tested": true,
20-
"notes": "Primary tested path."
35+
"notes": "Primary tested path. Ubuntu 22.04 and 24.04 are the validated distros; the installer's package-manager probes assume apt-get. Other distros (Fedora, Rocky, Alma, NixOS, Arch) may work but are not validated."
2136
},
2237
{
2338
"name": "macOS (Apple Silicon)",
@@ -125,25 +140,197 @@
125140
"name": "Local Ollama",
126141
"status": "caveated",
127142
"endpoint_type": "Local Ollama API",
128-
"notes": "Available when Ollama is installed or running on the host"
143+
"notes": "Available when Ollama is installed or running on the host. Validated default models: `qwen3.6:35b` (high VRAM), `nemotron-3-nano:30b` (medium VRAM), `qwen3.5:9b` (low VRAM fallback)."
129144
},
130145
{
131146
"name": "Local NVIDIA NIM",
132147
"status": "experimental",
133148
"endpoint_type": "Local OpenAI-compatible",
134-
"notes": "Requires `NEMOCLAW_EXPERIMENTAL=1` and a NIM-capable GPU"
149+
"notes": "Requires `NEMOCLAW_EXPERIMENTAL=1` and a NIM-capable NVIDIA GPU. Host must have the NVIDIA Container Toolkit installed and a CDI spec present (`onboard` asserts CDI presence). NIM images pull from `nvcr.io` and require NGC registry login. Validated images: `nvidia/nemotron-3-super-120b-a12b` (40 GB), `nvidia/nemotron-3-nano-30b-a3b` (8 GB), `nvidia/llama-3.1-nemotron-70b-instruct` (81 GB), `nvidia/llama-3.3-nemotron-super-49b-v1` (24 GB), `meta/llama-3.1-8b-instruct` (16 GB)."
135150
},
136151
{
137152
"name": "Local vLLM (already running)",
138153
"status": "caveated",
139154
"endpoint_type": "Local OpenAI-compatible",
140-
"notes": "Appears in the onboarding menu when NemoClaw detects a server already on `localhost:8000`. No flag required."
155+
"notes": "Appears in the onboarding menu when NemoClaw detects a server already on `localhost:8000`. No flag required. Model is whatever the existing server serves."
141156
},
142157
{
143158
"name": "Local vLLM (managed install/start)",
144159
"status": "caveated",
145160
"endpoint_type": "Local OpenAI-compatible",
146-
"notes": "Appears by default on DGX Spark and DGX Station. Generic Linux NVIDIA GPU hosts require `NEMOCLAW_EXPERIMENTAL=1` or `NEMOCLAW_PROVIDER=install-vllm`. NemoClaw pulls/starts a vLLM container on a supported NVIDIA GPU host."
161+
"notes": "Appears by default on DGX Spark and DGX Station. Generic Linux NVIDIA GPU hosts require `NEMOCLAW_EXPERIMENTAL=1` or `NEMOCLAW_PROVIDER=install-vllm`. Host must have the NVIDIA Container Toolkit installed and a CDI spec present (`onboard` asserts CDI presence). NemoClaw pulls/starts the stable NGC `nvcr.io/nvidia/vllm:26.05.post1-py3` container image on a supported NVIDIA GPU host. Validated defaults: DGX Spark → `nvidia/Qwen3.6-35B-A3B-NVFP4`; DGX Station → `Qwen/Qwen3.6-27B-FP8`; Linux NVIDIA GPU → `nvidia/NVIDIA-Nemotron-3-Nano-4B-FP8`. Image pulls require NGC registry login (`docker login nvcr.io`); onboard prompts for the NGC API key when authentication is missing."
162+
}
163+
],
164+
165+
"agents": [
166+
{
167+
"name": "OpenClaw",
168+
"status": "tested",
169+
"default": true,
170+
"notes": "Default agent runtime. Onboard with `$$nemoclaw onboard` (no `--agent` flag required)."
171+
},
172+
{
173+
"name": "Hermes",
174+
"status": "caveated",
175+
"default": false,
176+
"notes": "First-class agent with dedicated CLI (`$$nemohermes`), Dockerfile, manifest, docs, and nightly E2E coverage. Onboard with `$$nemohermes onboard` or pass `--agent hermes` to `$$nemoclaw onboard`. Unlocks the Hermes Provider inference route. Known limitations: model-provider compatibility registry is empty (backfilled per failure, see `nemoclaw-blueprint/model-specific-setup/hermes/README.md`); no Hermes-specific unit tests in `nemoclaw/src/`; macOS and WSL CI suites do not differentiate agents; deployment-identity bugs in the active release window (see #5211, #5327). Suitable for evaluation and the documented onboarding paths; production parity with OpenClaw is not yet asserted."
177+
}
178+
],
179+
180+
"capabilities": [
181+
{
182+
"name": "Guided onboarding",
183+
"status": "tested",
184+
"notes": "Single-command interactive wizard (`$$nemoclaw onboard`) that walks the user through inference provider selection, credential setup, sandbox creation, and dashboard launch. Non-interactive mode is supported via `--non-interactive` + `NEMOCLAW_*` environment variables for CI and scripted installs."
185+
},
186+
{
187+
"name": "Sandboxed execution",
188+
"status": "tested",
189+
"notes": "Landlock, seccomp, network namespace isolation, no-new-privileges, privilege dropping, and process limits (ulimit -u 512). Capability drops enforced for CAP_SYS_ADMIN, CAP_SYS_PTRACE, CAP_NET_RAW, CAP_DAC_OVERRIDE, CAP_SYS_CHROOT, CAP_FSETID, CAP_SETFCAP, CAP_MKNOD, CAP_AUDIT_WRITE, CAP_NET_BIND_SERVICE. Bounding-set tightening on hosts without CAP_SETPCAP is partial; see #3280."
190+
},
191+
{
192+
"name": "Routed inference",
193+
"status": "tested",
194+
"notes": "Provider-routed model calls through the OpenShell gateway, transparent to the agent. The agent uses `inference.local` inside the sandbox; provider credentials stay on the host. Supports every entry in the Providers table."
195+
},
196+
{
197+
"name": "Declarative network policy",
198+
"status": "tested",
199+
"notes": "YAML-defined egress with policy presets. Presets include `slack`, `discord`, `telegram`, `weather`, `openclaw-pricing`, `huggingface`, `npm`, `pypi`, `brew`, and others. Hot-reloadable at runtime via `$$nemoclaw <name> policy-add`."
200+
},
201+
{
202+
"name": "Snapshot and restore",
203+
"status": "tested",
204+
"notes": "Create, list, and restore named snapshots of sandbox state via `$$nemoclaw <name> snapshot create|list|restore`. Credential stripping is enforced on capture. Unsafe symlinks are rejected on restore."
205+
},
206+
{
207+
"name": "Agent skills",
208+
"status": "tested",
209+
"notes": "Packaged agent skills are discoverable by Cursor, Claude Code, and other coding assistants via `.agents/skills/`. Skills also install into the sandbox via `$$nemoclaw <name> skill install`."
210+
},
211+
{
212+
"name": "State migration",
213+
"status": "tested",
214+
"notes": "Sandbox state migrates across rebuilds with credentials intentionally excluded. Hermes excludes `auth.json` and restores its SQLite session DB through the backup API. OpenClaw config merge prevents stale state from overwriting fresh values."
215+
},
216+
{
217+
"name": "Blueprint versioning",
218+
"status": "tested",
219+
"notes": "Versioned, digest-verified, and reproducible blueprint lifecycle. Drives `$$nemoclaw <name> rebuild` and the migration safeguards above."
220+
},
221+
{
222+
"name": "Web search backend",
223+
"status": "caveated",
224+
"notes": "Runtime-configurable web-search backend (Brave, Tavily) plumbed through the OpenShell gateway. Backend credentials are user-supplied; no NemoClaw-bundled key. Backend availability depends on the chosen provider."
225+
}
226+
],
227+
228+
"out_of_scope": [
229+
{
230+
"name": "Podman / other container runtimes",
231+
"status": "unsupported",
232+
"notes": "Onboard rejects Podman explicitly (`src/lib/onboard.ts:1564` and `src/lib/onboard/preflight.ts:586`). Only Docker Engine, Docker Desktop, and Colima are supported. See issue #420 (closed)."
233+
},
234+
{
235+
"name": "Intel Mac (macOS x86_64)",
236+
"status": "unsupported",
237+
"notes": "OpenShell does not publish macOS x86_64 standalone gateway assets. Install hard-fails on x86_64 macOS (`scripts/install-openshell.sh:315`). See issue #954 (closed)."
238+
},
239+
{
240+
"name": "Non-Ubuntu/Debian Linux distros",
241+
"status": "unsupported",
242+
"notes": "Installer assumes `apt-get`. Fedora/Rocky/Alma/Arch/NixOS are not validated and the installer's package-manager probes do not cover them. See open issue #899 (Fedora hang)."
243+
},
244+
{
245+
"name": "Native Kubernetes or OpenShift deployments",
246+
"status": "unsupported",
247+
"notes": "NemoClaw runs the sandbox as a Docker container, not a Kubernetes pod. The default Docker-driver topology does not embed k3s. Operator-managed K8s/OpenShift deployments are out of scope; see issue #407 (community OpenShift via agent-sandbox CRD) and #2218 (epic)."
248+
},
249+
{
250+
"name": "Air-gapped / offline installs",
251+
"status": "unsupported",
252+
"notes": "Onboard assumes network reachability for package fetches, container pulls, and provider validation. See open issue #4872."
253+
},
254+
{
255+
"name": "Windows-on-ARM GPU passthrough",
256+
"status": "unsupported",
257+
"notes": "Windows-on-ARM CPU paths run under WSL2 'tested with limitations', but GPU passthrough on WOA is denylisted (`src/lib/onboard/wsl-docker-desktop-gpu.ts:188`, `src/lib/inference/gpu-trust.test.ts:70`). See closed issue #4565."
258+
},
259+
{
260+
"name": "Non-NVIDIA GPUs (AMD/ROCm, Intel Arc, Apple Metal)",
261+
"status": "unsupported",
262+
"notes": "Local vLLM and NIM paths assert NVIDIA CDI presence (`src/lib/onboard.ts:5041`). NemoClaw does not install non-NVIDIA accelerator drivers."
263+
},
264+
{
265+
"name": "LangChain, AutoGen, CrewAI, or other agent harnesses",
266+
"status": "unsupported",
267+
"notes": "Only OpenClaw and Hermes are integrated. Bringing other harnesses is tracked as a research epic (see open issue #4861) but is not on the current roadmap."
268+
},
269+
{
270+
"name": "Multi-user host sharing",
271+
"status": "unsupported",
272+
"notes": "Sandboxes are scoped to a single host user. NemoClaw treats multi-user hosts as a risk and warns at onboard; see `docs/security/openclaw-controls.mdx` Multi-user detection."
273+
},
274+
{
275+
"name": "Hosted SaaS / managed NemoClaw",
276+
"status": "unsupported",
277+
"notes": "There is no managed offering. Supported deployment paths are Local CLI onboard, Remote GPU via Brev CLI, and Brev web UI."
278+
},
279+
{
280+
"name": "Native provider integrations not in the Providers table",
281+
"status": "unsupported",
282+
"notes": "Vertex AI, Azure OpenAI, SageMaker, Together.ai, Replicate, and HuggingFace Inference Endpoints are not first-class onboarding entries. AWS Bedrock works through the `compatible-anthropic-endpoint` adapter (`src/lib/onboard/bedrock-runtime.ts`)."
283+
},
284+
{
285+
"name": "Production SLA or guaranteed response times",
286+
"status": "unsupported",
287+
"notes": "NemoClaw is an early-preview alpha project. Maintainers respond on a best-effort basis. No SLA is offered."
288+
}
289+
],
290+
291+
"integrations": [
292+
{
293+
"name": "Slack",
294+
"status": "tested",
295+
"notes": "Configured via OpenShell-managed channel during onboarding. Sandbox egress allowed by the `slack` policy preset."
296+
},
297+
{
298+
"name": "Discord",
299+
"status": "tested",
300+
"notes": "Configured via OpenShell-managed channel during onboarding. Sandbox egress allowed by the `discord` policy preset."
301+
},
302+
{
303+
"name": "Telegram",
304+
"status": "tested",
305+
"notes": "Configured via OpenShell-managed channel during onboarding."
306+
},
307+
{
308+
"name": "WeChat",
309+
"status": "caveated",
310+
"notes": "Channel hook available. Verify regional account access before relying on this path."
311+
},
312+
{
313+
"name": "WhatsApp",
314+
"status": "caveated",
315+
"notes": "Channel hook available. Verify Meta Business API access before relying on this path."
316+
}
317+
],
318+
319+
"deployment_paths": [
320+
{
321+
"name": "Local CLI onboard",
322+
"status": "tested",
323+
"notes": "Run `$$nemoclaw onboard` on a tested platform with Docker available locally. Primary path."
324+
},
325+
{
326+
"name": "Remote GPU via Brev CLI",
327+
"status": "caveated",
328+
"notes": "Legacy compatibility wrapper provisions a Brev VM, installs Docker + NVIDIA Container Toolkit, and runs `$$nemoclaw onboard` on that host. Defaults to GCP; override with `NEMOCLAW_BREV_PROVIDER`. The preferred path is the standard installer followed by `$$nemoclaw onboard` after the VM is reachable."
329+
},
330+
{
331+
"name": "Brev web UI",
332+
"status": "tested",
333+
"notes": "Browser-driven launcher provisions a Brev-managed Linux VM with Docker, the OpenShell runtime, a NemoClaw sandbox running OpenClaw, inference routing, and the OpenClaw dashboard."
147334
}
148335
]
149336
}

‎docs/about/overview.mdx‎

Lines changed: 2 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -22,6 +22,8 @@ NemoClaw pairs hosted models on inference providers or local endpoints with a ha
2222
The sandbox runtime comes from [NVIDIA OpenShell](https://github.com/NVIDIA/OpenShell).
2323
NemoClaw adds the blueprint, <AgentCli /> CLI, onboarding, and related tooling as the reference way to run supported agents there.
2424

25+
For the canonical list of what NemoClaw supports today across platforms, inference providers, agents, messaging integrations, and deployment paths, see [Platform Support and Launch Claims](../reference/platform-support).
26+
2527
| Capability | Description |
2628
|-------------------------|------------------------------------------------------------------------------------------------------------------------------------------------------|
2729
| Sandbox supported agents | Creates an OpenShell sandbox pre-configured for your selected agent, with filesystem and network policies applied from the first boot. |

‎docs/about/release-notes.mdx‎

Lines changed: 1 addition & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -14,6 +14,7 @@ import { AgentOnly } from "../_components/AgentGuide";
1414
NVIDIA NemoClaw is available in early preview starting March 16, 2026.
1515
Use this page to track the highlights of the latest release.
1616
For more detailed release notes, refer to the [NemoClaw GitHub announcements](https://github.com/NVIDIA/NemoClaw/discussions/categories/announcements?discussions_q=is%3Aopen+category%3AAnnouncements).
17+
For the canonical list of supported platforms, providers, agents, integrations, and deployment paths with explicit status per row, see [Platform Support and Launch Claims](../reference/platform-support).
1718

1819
## v0.0.63
1920

‎docs/get-started/prerequisites.mdx‎

Lines changed: 2 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -71,12 +71,13 @@ Refer to the [troubleshooting guide](../reference/troubleshooting) for the overr
7171

7272
The following table lists tested platform and runtime combinations.
7373
Availability is not limited to these entries, but untested configurations can have issues.
74+
For the full canonical matrix including deferred platforms, inference providers, agents, messaging integrations, and deployment paths, see [Platform Support and Launch Claims](../reference/platform-support).
7475
The table comes from [`ci/platform-matrix.json`](https://github.com/NVIDIA/NemoClaw/blob/main/ci/platform-matrix.json), the single source of truth kept in sync by CI and QA.
7576

7677
{/* platform-matrix:begin */}
7778
| OS | Container runtime | Status | Notes |
7879
|----|-------------------|--------|-------|
79-
| Linux | Docker | Tested | Primary tested path. |
80+
| Linux | Docker | Tested | Primary tested path. Ubuntu 22.04 and 24.04 are the validated distros; the installer's package-manager probes assume apt-get. Other distros (Fedora, Rocky, Alma, NixOS, Arch) may work but are not validated. |
8081
| macOS (Apple Silicon) | Colima, Docker Desktop | Tested with limitations | Install Xcode Command Line Tools (`xcode-select --install`) and start the runtime before running the installer. |
8182
| DGX Spark | Docker | Tested | Use the standard installer and `$$nemoclaw onboard`. For an end-to-end walkthrough with local Ollama inference, see the [NVIDIA Spark playbook](https://build.nvidia.com/spark/nemoclaw). |
8283
| Windows WSL2 | Docker Desktop (WSL backend) | Tested with limitations | Requires WSL2 with Docker Desktop backend. |

‎docs/index.mdx‎

Lines changed: 2 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -37,6 +37,8 @@ NVIDIA NemoClaw is an open-source reference stack for running always-on AI agent
3737
NemoClaw provides onboarding, lifecycle management, and agent operations within OpenShell containers.
3838
It installs the OpenShell runtime, part of NVIDIA Agent Toolkit, and sets up an environment designed for executing agents with additional security and inference routing capabilities.
3939

40+
For the canonical list of supported platforms, inference providers, agents, messaging integrations, and deployment paths with explicit status per row, see [Platform Support and Launch Claims](./reference/platform-support).
41+
4042
## Get Started
4143

4244
Install NemoClaw and run the onboard wizard to get started.

0 commit comments

Comments
 (0)