All plugins

agent-data-tools

Over 600 hosted data and AI tools through three MCP tools (find_tool, tool_details, call_tool), so the agent's context holds three definitions: web search, the text of a web page or PDF, screenshots, weather, FX, stock and token prices, wallet and chain reads, sanctions and company checks, DNS lookups, LLM chat, speech to text, converters and calculators. 25 free calls with no key or sign-up, and one bundled skill. Portable Agent Plugins v1 package (plugin.json + mcp.json + skills/), configuration only. Disclosures - Hosted-only connector with no local code: every tool call, with its arguments (the words given to find_tool; search words, URLs, addresses or text given to call_tool), goes to agent-data-tools.revmesh2074.workers.dev, which fetches named pages from the public web and sends each request to the public data source or model behind the tool that was called. - As installed it cannot spend money: no key is declared or stored, and after the free calls a tool answers with a payment request and stops. - A user who buys a prepaid key (from $1) adds their own same-name config.yaml entry with the key as a bearer header; each successful call_tool call then spends that key's credits at the called tool's price (1 to 100 credits, most often 10; 1 credit = $0.001). The package declares that range and a spend limit of 1,000 credits per chat; find_tool and tool_details are free. - The server is marked untrusted. All three tools are annotated read-only: this server cannot reach the service's tools that write. - The service keeps call counts per tool and day, a one-way hash of the caller's IP address for the free-call allowance, and its host's ordinary request logs. No telemetry, no background process, no self-update. - The tool set is served by the remote server at runtime and is not covered by the pinned sha.

Nutshell2000Toolsv1.0.1Updated 19 minutes ago

About

Over 600 hosted data and AI tools through three MCP tools (find_tool, tool_details, call_tool), so the agent's context holds three definitions: web search, the text of a web page or PDF, screenshots, weather, FX, stock and token prices, wallet and chain reads, sanctions and company checks, DNS lookups, LLM chat, speech to text, converters and calculators. 25 free calls with no key or sign-up, and one bundled skill. Portable Agent Plugins v1 package (plugin.json + mcp.json + skills/), configuration only. Disclosures - Hosted-only connector with no local code: every tool call, with its arguments (the words given to find_tool; search words, URLs, addresses or text given to call_tool), goes to agent-data-tools.revmesh2074.workers.dev, which fetches named pages from the public web and sends each request to the public data source or model behind the tool that was called. - As installed it cannot spend money: no key is declared or stored, and after the free calls a tool answers with a payment request and stops. - A user who buys a prepaid key (from $1) adds their own same-name config.yaml entry with the key as a bearer header; each successful call_tool call then spends that key's credits at the called tool's price (1 to 100 credits, most often 10; 1 credit = $0.001). The package declares that range and a spend limit of 1,000 credits per chat; find_tool and tool_details are free. - The server is marked untrusted. All three tools are annotated read-only: this server cannot reach the service's tools that write. - The service keeps call counts per tool and day, a one-way hash of the caller's IP address for the free-call allowance, and its host's ordinary request logs. No telemetry, no background process, no self-update. - The tool set is served by the remote server at runtime and is not covered by the pinned sha.

Agent Data Tools for Hermes Agent

A Portable Agent Plugins v1 package for Hermes Agent.
It gives a Hermes agent the catalog of Agent Data Tools, over
600 hosted tools, through three MCP tools, so the agent's context holds three tool definitions and not hundreds:

| Tool | What it does | Cost |
|---|---|---|
| find_tool | says which tools do a job described in plain words, with each one's price and inputs | free |
| tool_details | every input, the price and an example call of one tool | free |
| call_tool | runs a tool by name with its arguments | that tool's own price |

What the catalog covers: web search, the text of a web page or PDF, screenshots, weather, FX, stock, commodity
and token prices, wallet and chain reads on EVM networks, sanctions, company and VAT checks, DNS and domain lookups,
LLM chat, embeddings, speech to text and text to speech, and about a hundred converters, hashes and calculators.

The package holds one MCP server entry and one skill, and nothing else: no Python, no executables, no local command.

| Part | What it is |
|---|---|
| agent-data-tools | the hosted MCP server https://agent-data-tools.revmesh2074.workers.dev/mcp/compact-capped (Streamable HTTP, 3 tools) |
| skills/agent-data-tools | how to find and call a tool, how to treat what comes back, and what calls cost |

Install

hermes plugins install agent-data-tools
hermes plugins enable agent-data-tools

Portable packages install disabled, so the second command is needed. The tools arrive as
mcp__agent-data-tools__find_tool, mcp__agent-data-tools__tool_details and mcp__agent-data-tools__call_tool.

What it costs

  • Free to try. find_tool and tool_details are always free. The first 25 successful call_tool calls need no
    key and no sign-up, and failed calls are never counted. The free calls cover the tools find_tool marks
    inFreeCalls: reading pages and PDFs, prices, weather, chain reads, checks, lookups and the converters. Search,
    rendered pages, screenshots, speech, image and large-model tools cost us money per call and need a key from the first
    call.
  • Then pay per successful call. A prepaid key costs from $1 (1,000 credits; 1 credit = $0.001) at
    /credits, by card through Stripe, with no account. Most
    tools cost 10 credits a call, reading a page or a PDF 2, and a converter, hash or calculator 1. No tool this server
    reaches costs more than 100. Failed calls are refunded to the key at once. Credits bought as a pack do not expire.
  • What Hermes counts. Hermes knows call_tool by one name, so the package declares its cost as a range of 1 to 100
    credits, and Hermes books the top of a range for every call. With the package's limit of 1,000 credits ($1.00) per
    chat, Hermes therefore asks after ten call_tool calls in one chat, although they usually spent far less. Your key is
    only ever charged each tool's real price. Change the limit with spend_limit in your own config.yaml entry.

Use a key

The package itself carries no key (Hermes will not start a server whose header names a variable that is not set, and
the free calls should work without one). When you have a key, put it in your profile .env as
AGENT_DATA_TOOLS_KEY=adt_... and add this to ~/.hermes/config.yaml, then run /reload-mcp:

mcp_servers:
  agent-data-tools:
    url: "https://agent-data-tools.revmesh2074.workers.dev/mcp/compact-capped"
    headers:
      Authorization: "Bearer ${AGENT_DATA_TOOLS_KEY}"

An entry of the same name in config.yaml replaces the package's server entry and keeps the costs it declares. The
same file is in this folder as config.yaml.

Disclosure

  • Network. Every tool call goes to agent-data-tools.revmesh2074.workers.dev with its arguments: the words given
    to find_tool, and for call_tool the tool's inputs (search words, URLs, addresses, text to process). The service
    fetches the pages and files you name from the public web on your behalf and sends each request to the public data
    source or model behind the tool that was called. Nothing else on your machine is read or sent.
  • Money. Without a key no call can spend anything: after the free calls a tool answers with a payment request and
    stops. With your key in config.yaml, each successful call_tool call spends that key's prepaid credits at the
    called tool's price, never more than 100 credits ($0.10) a call. The plugin cannot buy credits and holds no card.
  • Approvals. The package marks the server untrusted. All three tools are annotated read-only: this server cannot
    reach the service's four tools that write (saved agent memory and page publishing), so nothing it runs changes
    anything on your machine or in your accounts. Hermes stops and asks when a chat reaches the spend limit.
  • Credentials. The package declares none. A key you add lives in your own profile .env and is sent as a bearer
    header by Hermes's own MCP client.
  • What the service keeps. Counts of calls per tool and per day (no inputs or outputs in them), a one-way hash of
    the caller's IP address for the free-call allowance, and for a key a one-way hash of it with its balance and
    purchase history. Its host, Cloudflare, keeps ordinary request logs (IP address, user agent, URL, timing) for days.
    Privacy ·
    Terms.
  • No telemetry, no background process, no self-update. The tool set is served by the remote server at runtime and
    is not covered by the pinned commit.

Tools above $0.10 a call

This package connects the capped address on purpose: there, a tool priced above $0.10 a call does not exist, so the
100 credits Hermes books per call is a real upper limit. The same three tools without the cap are at /mcp/compact,
where call_tool runs a dearer tool when the call names its price in max_price. If you connect that address
yourself under this server's name, Hermes still books 100 credits for such a call, so its count can then be too low.

One section instead

The same service serves each section of the catalog as an ordinary MCP server with one definition per tool, priced
the same way: /mcp/web, /mcp/ai, /mcp/everyday, /mcp/prices, /mcp/chain, /mcp/compliance and
/mcp/utilities. Add one under another name in config.yaml if you would rather have its tools listed by name; this
package declares no costs for those, so Hermes will not count their spend.

Source of this package: https://github.com/Nutshell2000/agent-data-tools/tree/main/integrations/hermes. Support:
https://agent-data-tools.revmesh2074.workers.dev/support.

README from the reviewed commit c3b9c59; it updates when the author re-pins.

Developer Mode

Explore built-in Hermes toolsets and actions.

HermesAgent

Developer?

Create a Hermes extension for Desktop or agent workflows.

Developer Section