hypragent-registry

HyprAgent MCP Registry

Curated list of Model Context Protocol servers that HyprAgent’s “Tools → Registry” tab can install in one click.

The registry is opt-in in HyprAgent (since 0.5.6) — the daemon only fetches this file when the user explicitly enables it in Settings, after acknowledging a consent dialog explaining the network call.

Where this is served

The canonical URL the daemon fetches is:

https://registry.hypragent.dev/registry.toml

This repo’s main branch is published via GitHub Pages, with a custom domain mapped onto registry.hypragent.dev (a dedicated subdomain so the registry can grow independently of any future content on the apex). The daemon fetches with If-None-Match against the previous ETag, caches locally for 24h, and falls back to the cached copy when the network is unavailable.

What the daemon does with this file

For each [[servers]] entry, HyprAgent shows:

• A card in the Registry tab with the icon, name, and description.
• An Install button that opens the existing MCP server form pre-populated with the entry’s command / args / env.

The user fills in any env_required values (with secrets routed to secret.toml via the per-key secret = true flag) and clicks Save. HyprAgent writes a normal [mcp.<id>] block to the user’s config; the third-party server then runs as a separate process under the user’s account.

HyprAgent does not audit, sandbox, or run these servers. The registry is a curated list of canonical configurations — nothing more. The trust model is the same as npm install <package> followed by node ./bin/<package>: you are running upstream code with your credentials.

Schema

[[servers]]
id                  = "github"             # slug, becomes the [mcp.<id>] key
name                = "GitHub"             # human-readable label
description         = "..."                # one-line, shown in the card
icon                = "🐙"                 # emoji or absolute image URL
homepage            = "https://..."        # upstream project / docs
type                = "stdio"              # "stdio" | "sse" | "http"
command             = "npx"                # stdio only
args                = ["-y", "..."]        # stdio only
url                 = ""                   # sse / http only
default_permission  = "ask"                # "allow" | "ask" | "deny" applied
                                           # to every mcp__<id>__* tool
signature           = ""                   # reserved for future ed25519 signing

env_required = [
    {name = "GITHUB_PERSONAL_ACCESS_TOKEN", description = "...", secret = true},
]
env_optional = []

env_required / env_optional entries:

• name — environment variable name.
• description — shown to the user in the install dialog.
• secret — true → value written to secret.toml [mcp.<id>] (mode 0600). false → written to config.toml [mcp.<id>.env].

Adding or updating a server

1. Open a PR against this repo with the new [[servers]] entry.
2. Verify locally with the daemon: point your ~/.config/hypragent/ config.toml at a local copy via [registry] url = "file:///path/to/your/registry.toml", restart the daemon, open the Registry tab.
3. PR description should explain why the server is worth curating: real-world utility, upstream maintainership, no surprising network behaviour beyond what the description says.

Servers that have not been updated in 12+ months by their upstream maintainers will be flagged for removal. The list is not a graveyard.

Reporting issues with a listed server

If a registry entry is broken (wrong command, env-var schema drift upstream, server itself misbehaves), open an issue here and include:

• Which entry (id).
• What you observed (error message, hang, wrong behaviour).
• Whether the upstream project’s own README still matches what the registry says.

If the issue is with HyprAgent’s handling of a registry entry (install dialog crashes, env-vars not routed correctly), open the issue in the HyprAgent repo instead.