yourspace
docshome
blog

Indie Agents Welcome

The web deserves more wholesome content. If you're building an agent that generates blogs, docs, portfolios, or community sites — the deploy step should be the smallest part of your pipeline, not the part that demands a dashboard tab.

The idea

Your agent does the creative work. YourSpace handles the last mile:

  1. Build — generate HTML, markdown, whatever
  2. Bundle — tar -czf dist.tar.gz -C out .
  3. Deploy — one API call or MCP tool invocation

No dashboards. No OAuth flows. A token and a tarball; both fit in a single API call.

REST API

The simplest path. Generate a scoped token, then POST your bundle:

curl -X POST https://api.yo.urspace.net/v1/sites \
  -H "Authorization: Bearer $YOURSPACE_TOKEN" \
  -F "bundle=@dist.tar.gz"

The response carries the URL and the version number the deploy minted. Same endpoint, same shape, whether the caller is a human at a terminal or an agent in a loop.

MCP integration (design preview)

If your agent speaks MCP, the plan is to expose the same surface through a hosted server at mcp.yo.urspace.net. The client-side config is the usual shape:

{
  "mcpServers": {
    "yourspace": {
      "url": "https://mcp.yo.urspace.net",
      "headers": {
        "Authorization": "Bearer $YOURSPACE_TOKEN"
      }
    }
  }
}

The agent gets five tools: yourspace_deploy, yourspace_list_sites, yourspace_get_site, yourspace_delete_site, and yourspace_create_token. That's the entire surface area — and it'll stay that size; MCP is a layer over the same REST endpoints, not a second API to keep in sync.

Why MCP?

MCP is becoming the standard way for AI agents to interact with external services. Instead of teaching every agent how to shell out to a CLI or construct curl commands, MCP gives them a typed, discoverable tool interface.

An agent using our MCP server will be able to:

  • Deploy a freshly generated static site
  • Check when a deploy has gone live
  • Tear down old versions after confirming the new one is live
  • Mint scoped tokens for sub-agents or CI pipelines

All without knowing anything about HTTP, tar, or DNS.

Token shape for agents (where this is heading)

  • Scoped tokens are the right shape — an agent's token should carry only the verbs it needs (deploy, list, read — not delete, not mint). The current token is still a bearer for the whole account; scoping is coming alongside the auth buildout.
  • Rotation should be cheap. yourspace_create_token exists in the design so an orchestrator can mint a short-lived token per job and discard it after.
  • Bundle integrity. Uploads are validated on the server before anything is written to disk; corrupted tarballs fail fast.

What people are building

  • AI-generated blogs — an agent writes a post, builds it with Hugo, and deploys via MCP
  • Documentation bots — watch a repo, rebuild docs on every commit, push to edge
  • Portfolio generators — scrape a user's projects, generate a static portfolio, deploy it under their domain

The pattern is always the same: generate → bundle → deploy. YourSpace stays out of the way.

Build the agent. We'll handle the edge.