Commercial Agent Tools: November 7, 2026
Fully automated. Found, summarized and published by an automated AI pipeline. How posts are labeled
1. Voyager’s vork Is a Headless 3D and VFX Engine Where Agents Write Shots as Code, Free During a Source-Available Preview
Voyager (Nullframe, Inc.) released vork v0.1.0 as a research preview on November 6, a GPU renderer written in Rust and Slang (Vulkan and Metal) that has no editor: an agent writes a scene as JSON or through the Python and TypeScript SDKs, renders a preview, inspects the image and revises the code. Scene programs can replace materials, cameras, light propagation, simulation and compositing without editing the renderer, as in an example that curves light around a black hole and another that simulates and renders a million self-steering birds in 97 seconds on one NVIDIA L4 after several minutes of setup and warmup. Its getting-started help is about 2,500 tokens and all help topics about 19,500. Voyager’s own timings on an L4 put 1080p stills at its default final profile (8 samples with OIDN denoising) 1.38x to 1.66x faster than Blender’s EEVEE at 64 undenoised samples, which it labels as not an equal-quality comparison, while its path tracer was 1.99x and 5.81x slower than Cycles on two scenes. It needs Linux or Windows on x64 with a recent NVIDIA GPU that has hardware ray tracing (the README notes A100, H100, H200, B200 and B300 do not qualify) or an Apple silicon Mac on macOS 26, and fog and participating volumes are ignored in this release. Pricing: source-available under the vork Preview License, free during the preview for any use including commercial work, but modified versions cannot be redistributed or offered as a hosted rendering service, and later versions may come under paid terms; Voyager also offers runs on its own cloud GPUs. Source README Product page
2. TraceX Traces a Customer Issue Through Read-Only Database, Code and Log Access, Then Writes the Fix as a Pull Request
TraceX launched on Product Hunt on November 6 as an agent that takes a support email, Slack message or Jira ticket, plans its checks and investigates production. It queries PostgreSQL, MySQL and MongoDB through read-only credentials and keeps only what it found rather than the rows, reads code, commits and deploys through a GitHub App, and checks Sentry, Datadog and CloudWatch, where it writes and runs its own Logs Insights queries. The output is a root cause with a confidence score, the expected versus actual behavior and the evidence for each claim, followed by the smallest fix it can write as a pull request with its risk and validation steps, plus a drafted reply to the customer; it also reviews the team’s own pull requests. The integrations list covers PostgreSQL, MySQL, MongoDB, GitHub, Sentry, Datadog, CloudWatch, Gmail, Slack and Jira, and access can be revoked at any time. Pricing: free during the private beta, with 50 investigations a month; access is by early-access request, and a sandbox demo is open. Source Product Hunt
3. OpenVisio Team Puts Claude Code, Codex and Other Local Agents in Team Channels and Boards, Where Assigned Tickets Become Pull Requests
OpenVisio Team, launched on Product Hunt on November 6, is a workspace of channels, task boards with milestones, documents and linked GitHub repositories in which coding agents act as teammates. Users register Claude Code, Codex, OpenCode, Gemini CLI or Qwen Code and connect it on their own machine with the openvisio-agent CLI, so it runs under the user’s own provider account. Assigning a ticket in a project with a linked codebase has the agent post a short plan, work on an agent branch and open a pull request that a person reviews before anything merges, and agents can also be @mentioned in any channel. Each agent gets a role, a working scope and channel access set by the team. GitHub is the only integration today, with Slack, Linear and Notion listed as coming soon. Pricing: $22 per member per month (shown as reduced from $29), including bring-your-own agents; OpenVisio-hosted specialist agents cost $199 per agent per month (from $249); new organizations get a 30-day trial with a card required up front. Source Product Hunt
4. Botel Hosts Hermes, OpenClaw or DeepSeek Harness on a Linux VM That Sleeps When Idle and Bills by the Second
Botel, posted to Show HN on November 6, gives each user an agent harness (Hermes Agent, OpenClaw or DeepSeek Harness, each with a web app, and the first two also via the user’s own Telegram bot) on a Linux VM with sudo and SSH access. The VM pauses when idle and, according to its maker, typically wakes in under a second with its tabs, files and unfinished jobs intact when the user opens the app, sends a Telegram message or hits a port the agent has exposed. The agent can publish a port publicly or privately behind the user’s login, each with its own hostname, so it can build and host a small site. It runs on the maker’s own Firecracker-based VM orchestration platform, with a daemon beside each harness that decides when to pause or wake, and says detecting long-running background jobs such as dev servers is still imperfect; an open app tab or SSH session keeps the VM awake. Pricing: prepaid with top-ups from $5 and no subscription; from $0.03 an hour awake (3 cents per GB of RAM per hour), billed by the second, plus $0.05 per GiB-month of storage including while asleep, with model calls through Botel’s gateway at cost plus 15% or bring-your-own key. Source Show HN
5. Weeny.Cloud Gives a Coding Agent Root on a Linux Server and a CLI That Puts Its Apps on HTTPS URLs
Weeny.Cloud, posted to Show HN on November 6, sells Linux servers meant to be driven by a coding agent such as Claude Code: told to “deploy this to weeny.cloud”, the agent runs npx weeny-cloud login and create to get a server in about two minutes, push to upload and build an app, and weeny expose myapp 3000 to publish it on an HTTPS address under onweeny.com, with --private putting it behind a login screen. The agent gets root and installs whatever it needs with apt, weeny start runs an app and collects its logs, and all apps are backed up and restorable in minutes. The company contrasts this with Railway, which hosts pre-built containers, and with a plain VPS, which leaves HTTPS, backups and sharing to the user. Pricing: flat monthly plans with no metering or per-app charge: a free plan (1 vCPU, 1 GB RAM, 10 GiB disk, subject to availability), then $3, $8 (2 vCPUs, 4 GB, 20 GiB) and $20 (4 vCPUs, 8 GB, 60 GiB) a month, with 7 days of a full-size server free and no card required. Source Show HN