Sign inGet started

Live · 527 tools

Connect your agent to every AI tool it needs.

Give this to your agent

set up https://spiral.ai/SKILL.md

and let it take it from there.

Get your API key

Agent

Claude Codeagent:social
  1. find this week's formats
  2. generate the visuals
  3. cut the short-form edits
  4. scheduled the content plan

$ spiral run fal/openai/gpt-image-2.5/sunburst/edit ✓ completed

Social media toolbelt

+2,621 more tools

one key · pay per run

One key. Every tool your agent reaches for.

Spiral

Tell your agent what to do. It picks the tools itself.

  1. Discover and compare

    discover()

    every tool with its schema
    and its price

  2. Run the tool

    run()

    one call for any provider
    no per-provider sign-up

  3. Pay per run

    inspect()

    the price before every run
    from one balance

No subscriptions.

Your agent has one balance for every tool.

Tools across 17 providers
3,606
Pay only for the runs your agent makes
Per run
For every tool, every provider
1 balance

Three ways to connect.

One catalog and one balance, whichever your agent speaks.

  • SkillPaste one line into any agent. It reads the skill and sets itself up.
    set up https://spiral.ai/SKILL.md
    Read the docs
  • MCPAdd the hosted MCP server to your agent.
    claude mcp add spiral --transport http https://api.spiral.ai/mcp \
  --header "Authorization: Bearer $SPIRAL_API_KEY"
    Read the docs
  • APIOne REST endpoint and a webhook for every tool.
    curl "https://api.spiral.ai/v1/tools?limit=5" \
  -H "Authorization: Bearer $SPIRAL_API_KEY"
    Read the docs

Let your agent find the tools for you.

Pay as you go. Credits never expire.

set up https://spiral.ai/SKILL.md
Get your API key

Common questions.

Do I need an account with each provider?

No. One API key runs every tool in the catalog. We hold the provider accounts, so there is nothing else to sign up for.

How is pricing structured?

Pay as you go from one prepaid balance. Every tool lists how its price works, and every run shows what it was charged. There are no plans or minimums.

A run failed. Am I charged?

Not for a failure on our side or the provider's: the run says “You were not charged” and the hold goes back to your balance.

Which agents can use it?

Any MCP client, including Claude Code, Cursor, Claude Desktop and VS Code, plus anything that can call a REST API.

Can I cap what an agent spends?

Yes. Give each agent its own API key with a spend cap. The cap is reserved before a run starts, so it is never overshot.