Connect your agent to every tool it needs.
One balance, one line of setup. Your agent discovers a tool, checks the price, and runs it — paying per call, no sign-ups.
set up https://sooma.dev/SKILL.mdPaste that into Claude Code, Cursor, Codex or any MCP client.
Tell your agent what to do. It picks the tools.
Three verbs, in order. Nothing runs until your agent has seen the price.
sooma.discover({
query: "emails at stripe.com"
})sooma.inspect({
provider: "hunterio",
endpoint: "/domain-search"
})sooma.run({
provider: "hunterio",
endpoint: "/domain-search",
input: { domain: "stripe.com" }
})We killed the subscriptions.
- 17 tools across 13 providers, no sign-ups.
- Pay per call. The price is shown before every run.
- One balance for every tool. Provider errors cost nothing.
- Daily and per-call budgets when you want a hard ceiling.
Three ways to connect.
One line in your agent's chat. It reads the skill and knows the loop.
A remote MCP server with three tools: discover, inspect, run.
Plain HTTPS with a bearer key, from any language.
Let your agent find the tools for you.
Start with $1 in free credit.
set up https://sooma.dev/SKILL.mdCommon questions.
Do I need API keys for each provider?+
No. Sooma holds the vendor relationships. Your agent uses one Sooma key and one balance for every tool in the catalog.
How is pricing structured?+
Every endpoint lists its price up front — per call, or per returned row plus a small flat fee. inspect returns the exact rate before your agent runs anything.
A provider failed. Do I get charged?+
No. If the upstream returns an error, or Sooma itself fails, the run costs $0. Only successful responses are debited.
Which agents are supported?+
Anything that can read a skill file or speak MCP: Claude Code, Cursor, Codex, OpenCode, Claude.ai and plain HTTP clients.
Can I cap what an agent spends?+
Yes. Set a daily budget and a per-call limit in the dashboard; calls that would breach them are refused before they run.