Free tools for AI agents
Give your AI assistant the SBD planning tools
Connect Claude, ChatGPT, Codex, Cursor or any MCP client to the same six tools on this site. Agents get the same numbers a person gets on the web page. There is no login or key, every tool is read-only, and nothing you send is stored.
https://smallbaydigital.com/api/mcp Connect your client
Run in your terminal. Add --scope user to make it available in every project.
claude mcp add --transport http sbd-tools https://smallbaydigital.com/api/mcp
Settings, then Connectors, then Add custom connector. Name it SBD tools and paste the URL. Leave authentication empty.
https://smallbaydigital.com/api/mcp
Settings, then Apps and Connectors, then Advanced, and turn on Developer mode. Create a connector with this URL and no authentication.
https://smallbaydigital.com/api/mcp
Add to ~/.codex/config.toml. Older versions without remote server support can use the mcp-remote bridge shown under Any other client.
[mcp_servers.sbd-tools] url = "https://smallbaydigital.com/api/mcp"
Use the install button, or add to ~/.cursor/mcp.json.
{
"mcpServers": {
"sbd-tools": {
"url": "https://smallbaydigital.com/api/mcp"
}
}
} Use the install button, or add to .vscode/mcp.json in your project.
{
"servers": {
"sbd-tools": {
"type": "http",
"url": "https://smallbaydigital.com/api/mcp"
}
}
} Add to ~/.codeium/windsurf/mcp_config.json.
{
"mcpServers": {
"sbd-tools": {
"serverUrl": "https://smallbaydigital.com/api/mcp"
}
}
} Add to ~/.gemini/settings.json.
{
"mcpServers": {
"sbd-tools": {
"httpUrl": "https://smallbaydigital.com/api/mcp"
}
}
} Clients that only run local servers can reach it through the mcp-remote bridge. Needs Node.js.
{
"mcpServers": {
"sbd-tools": {
"command": "npx",
"args": [
"-y",
"mcp-remote",
"https://smallbaydigital.com/api/mcp"
]
}
}
} Five tools and a helper
automation_opportunity_mapOpportunity map on the web You are not sure what to automate. Ranks processes into waves with hours and money saved a year. "Map automation opportunities for our finance team: invoice entry 40 hours a month, payment chasing 25, month-end reconciliation 30." automation_roiROI calculator on the web One process is in view and you want to know if it pays. NPV, payback, IRR and the breakeven values. "Is automating 1,200 invoices a month worth it if each takes 6 minutes and the build costs $15,000?" build_vs_buyBuild vs buy on the web Choosing between buying software, configuring a platform or building. Three-year cost and weighted scores. "Should we buy a $30 a seat tool for 25 people or build our own for $40,000? We grow about 15% a year." ai_readinessAI readiness on the web Checking whether the organisation is ready for AI. Stage, pillar scores and next steps. "Run the AI readiness assessment for a 40-person company. Ask me the questions a few at a time." agent_riskAgent risk on the web A specific AI agent is planned. Risk tier, autonomy level, controls mapped to NIST, ISO 42001, the EU AI Act and OWASP, and a draft policy. "Assess the risk of an agent that reads customer emails and drafts refunds in our billing system." stack_checkStack checker on the web You have a company domain and want to know what software it runs or whether its email authentication is set up. Reads public DNS only. "What does the DNS for zapier.com say about the tools they use, and is their DMARC set up?" describe_tool Before running a tool. Lists every input with its unit, default, source, how much it matters and the question to ask. "What do you need from me to run the build vs buy tool?" A common path is opportunity map, then ROI on the top item, then build vs buy, then agent risk before launch. The server also offers a guided interview prompt for each tool and the full method of each as a resource.
How agents get good answers from partial information
- Start with what you know
Call a tool with the two or three numbers you have. Missing inputs use documented defaults, and the result lists every one it used.
- Read the confidence
Each result says high, medium or low confidence, based on how many of the inputs that move the answer most were given.
- Ask the questions it returns
Results include the next questions to ask, highest impact first. Ask them, then run the tool again.
- Plan an interview
describe_tool lists every input with its unit, range, default, source and the plain question to ask a person.
- Lead with the answer
Give the conclusion in one sentence, then the numbers behind it, then the assumptions that came from defaults.
- Hand over a link
Every result has a share link that opens the same inputs on the web page, where the person can adjust them and download a PDF.
Details
- Transport
- Streamable HTTP, stateless. JSON-RPC over POST to
https://smallbaydigital.com/api/mcp. Protocol versions 2025-06-18, 2025-03-26 and 2024-11-05. - Authentication
- None. The tools are public and read-only.
- Data
- Inputs are used to compute the answer and are not stored or logged with the inputs. The email-gated PDF on the web pages is separate and optional.
- Results
- Plain-text summary plus structured content: summary, confidence, defaulted, warnings, questions, input, result and share_link.
- Method
- Each tool page publishes its method and sources. Agents can also read them as MCP resources or as Markdown at /tools/<tool>.md.
- Not advice
- Results are estimates to decide with. They are not financial or legal advice.
Want someone to check the numbers or build it?
On a discovery call we go through what the tools returned for your team and tell you which result is worth acting on first.
Book a discovery call














