Your morning brief, your running log, the plan you asked for. Your AI agent publishes each one to a private link that opens anywhere. Comment on the page, and your agent fixes it on the same link.
toss report.htmlWatch the whole loop, step by step ↓
tossithere is a CLI and a skill file. Anything that can run a shell command can publish a page, read the comments and push v2.
Ask your agent for a morning briefing, a running log, a trip plan. If it can write it as HTML, it becomes a page on your phone.
Send me a briefing page every morning at 7.
Ran 5k this morning, 26:10
Plan 4 days in Lisbon in October, under $1,500.
Where did my money go in September?
Your agent can write the page. It can’t put it online, collect comments from three people, or keep track of which version they were about. That’s what tossithere does.
One command puts an HTML file and everything next to it online, private to you. It opens on your laptop or your phone, with no build step and nothing to host.
Click an element or highlight a run of text, and type. Each comment quotes what it is about, so your agent gets “the 12% in the growth card”, not “that number near the top”.
Revenue reached $4.2M, and growth was 12% QoQ, led by the Teams plan.
Toss the file again and the same link shows v2. Each version keeps its own comments, so you can see what changed and why.
Every doc starts private. Share it with the people you name, or anyone with the link, and choose whether they can read, comment or edit.
The writing stays with your agent. The link, the comments and every version live here.
It plays on its own. Click any step to jump to it.
| Oct 6 | Yearly option in checkout |
| Oct 20 | Email monthly subscribers |
| Nov 3 | Yearly becomes the default |
To be decided.
Anyone who reads what their agent writes, and everyone they send it to.
Read what your agent writes as a real page, on any screen. Comment instead of re-explaining, and your agent fixes it.
Share with anyone, even outside your org and on a different agent. Their comments land back in your agent.
Give your agent a review step without building comments, sharing and versions yourself. The CLI answers every command in JSON, with one error contract.
Or skip the terminal: paste one block of instructions and your agent sets itself up.
Your agent installs the CLI, signs in and publishes on its own.
Anyone you share a link with can read it, and comment after a Google sign-in, on every plan, forever. You pay for more documents, or to publish together as a team. One flat price, never per seat.
A shared workspace for everyone who publishes. Unlimited members and documents at one flat price, never per seat.
Start a teamRun the whole thing on your own infrastructure. Same CLI, same overlay, AGPL-3.0. Not released yet.
Request early accessThat works on a laptop. On a phone, especially an iPhone, an HTML file opens as a bare preview at best, without its scripts or images. Then you explain what's wrong in chat, and do it all again for every version. tossithere gives the file a link that opens on any screen, comments pinned to the exact spot, and v2 on the same link.
Some do, each in its own corner. If you use more than one agent, your pages end up in as many places. Some can't be shared with your team, most don't take comments, and the few that do keep them inside their own app. tossithere is one place for every agent's pages: the same links, the same sharing and the same comments, whichever agent made the page, and the comments go back to it.
For one person and one round, sure. Once two people review three versions, nobody knows which paragraph “this bit” meant. Every comment here quotes the element it's about, so your agent does.
Any agent that can run a shell command: Claude Code, Codex, Cursor, Cowork, OpenClaw, Hermes and more. Paste one block of instructions and it installs the CLI and publishes on its own.
Yes. Docs are private by default. Read them anywhere, on your laptop or your phone, comment on them, and your agent picks the comments up.
Viewing and commenting are free on every plan. Publishing is free for up to 10 documents. Pro is $10 a month for unlimited documents, and Team is $39 a month flat for a shared workspace.
Any HTML file and the assets next to it: reports, plans, specs, dashboards, prototypes. It is served as a static site, so it loads like one.
Only you, until you or your agent share one. Anyone with a shared link can read it. Commenting needs a Google sign-in, so your agent always knows who said what.
No. Every document is hidden from search engines by default. A public document has a switch if you do want it indexed.
Not yet. Documents run in a strict sandbox with outbound network access switched off, which is what makes it safe to host any HTML. Styles, fonts and common CDN libraries all work.
Each version keeps its own comments. Older versions stay open with their threads intact, so you can see what changed and why.
Pro is for one person: your own workspace, with no limit on documents. Team is a shared workspace where everyone in it can publish and manage documents, for $39 a month however many people join. A Team plan covers the team's workspace, and your personal workspace keeps its own plan.
No. It renders in a shadow root and floats above the page, so it never inherits your CSS or shifts your layout.
Agents write a lot of HTML now: plans, reports, briefings, dashboards. Getting it in front of people, and getting their feedback back to the agent, still happened in a chat box.
Claude artifacts are great, and they recently added comments. But they aren’t open to everyone, and they don’t plug into OpenClaw, Hermes or whatever other agent you might be running.
So I built tossithere: any agent can toss a page here, anyone you share it with can comment, and the comments go back to the agent.
Kush Daga@KushDaga · building tossithereKush Daga@KushDagaupdate on tossithere.comThe server, the viewer overlay and the CLI will be released together under AGPL-3.0, so you can run the whole thing behind your own VPN. It is not public yet. Want it sooner? Ask, and early access is a reply away.
Comment on the page instead of re-explaining it in chat. Your agent picks up every comment.
