Monid with Atomic Agent
Monid is a gateway to 3,100+ live data tools from 100+ providers: X and Reddit posts, leads and contacts, company and funding data, SEO, page screenshots, what ChatGPT says about a brand, even video generation. You top up one wallet and pay per call, with no subscriptions. This guide sets it up in Atomic Agent through Monid’s skill and the monid command line tool, with one API key. To connect it as an MCP server with one-click sign-in instead, see Monid MCP server.
How it works
Every Monid task follows three steps:
- Discover (free). The agent searches Monid’s catalog in plain words, for example “twitter posts by keyword”, and picks a tool.
- Inspect (free). It reads the tool’s input schema and price.
- Run (billed). It runs the tool, waits for slow jobs such as video to finish, and reports the result. The run record shows the exact cost.
Each step is a monid command the agent runs through os.shell.run. The key is stored by the monid tool on your computer, so it does not appear in the command text.
Setup
You need a Monid account, Atomic Agent (desktop app or terminal) and Node.js with npm, which installs the monid tool.
-
Create a key. Sign in at app.monid.ai, open Access › API keys and create one. It starts with
monid_live_. Add credit to the wallet while you are there. -
Install the Monid skill. Send Atomic Agent this line:
set up https://monid.ai/SKILL.mdThe agent reads Monid’s setup instructions, installs the
monidtool withnpmand saves the skill to your skills folder. Approve the commands when it asks. -
Add your key. The agent offers to save the key for you if you paste it. To keep the key out of the chat history, run this in your own terminal instead:
Terminal window monid keys add -k monid_live_YOUR_KEY -l main -
Ask. Start a request with “Using Monid, …” so the agent picks the skill.
Prefer to install by hand? Run these two commands, then restart Atomic Agent and save Monid’s SKILL.md as ~/.atomic-agent/skills/monid/SKILL.md:
npm install -g @monid-ai/climonid keys add -k monid_live_YOUR_KEY -l mainmonid keys list shows which key is active. See Skills for how skills load, and Monid’s docs for the full command list.
Prefer MCP?
Monid is also in Atomic Agent’s MCP server catalog (v0.8.0, Desktop v0.0.9 and later). Add it from Settings › Connections › MCP servers › Add server, or /mcp add in the terminal, and sign in with your Monid account. No API key and no npm install, and the Monid tools stay in the agent’s toolbox all the time. See Monid MCP server.
What to ask the agent
These ran in Atomic Agent on 16 and 17 Nov 2026. The costs are the amounts Monid charged, from its run records.
| Prompt | Monid tools | Cost |
|---|---|---|
| ”Using Monid, brief me on Linear before my meeting tomorrow: funding, what people say on X and Reddit, their pricing page, and what ChatGPT says about them.” | 5: company and funding data, X search, Reddit search, page screenshot, ChatGPT answer | $0.063 |
| ”Using Monid, turn this image into a 5-second promo clip: https://atomicagent.io/og.png” | 1: image-to-video | $0.18 |
| ”Using Monid, find 5 SaaS founders in San Francisco.” | 1: people search | $0.10 |
| ”Using Monid, find the top 5 posts on X about Monid this month and tell me what they say.” | 1: X search | $0.00049 |
The first prompt is the useful pattern: one request, several data sources, one written answer. The agent split it into five Monid calls, combined the results into a brief, flagged off-topic X results, and said which figures were the data provider’s claims rather than verified.
Some prices from tools we used:
| Data | Price |
|---|---|
| X search, 20 posts | about $0.0005 |
| Reddit search | about $0.0004 |
| Full-page screenshot | about $0.0013 |
| What ChatGPT says about a topic | about $0.0007 |
| Company and funding profile | $0.06 |
| Person profile in a people search | $0.02 each |
| Keywords a site ranks for | $0.012 plus $0.00012 per keyword |
| Video from an image, 768p | $0.036 per second |
Prices are set per tool by Monid and can change. monid inspect shows the current price of any tool, and you can ask the agent to check it before a run.
Spend and approvals
- Approvals on Atomic Agent’s side. Each Monid call is a shell command, so it goes through the same approval rules as any other command. At the default level the agent asks before each one, which is a good way to learn what calls cost. See Approval Gates.
- Limits on Monid’s side. Workspace budgets and per-run caps are set in Monid. A run over a limit comes back
BLOCKEDinstead of charging. - Balance.
monid balanceshows what is left in the wallet. - Keep runs small. Many tools charge per result. Ask for 10 to 20 results first, and more once you like what comes back.
- Emails cost extra. People searches return profiles without contact details. Revealing a work email costs more per person, so ask for it only when you need it.
Troubleshooting
monid: command not found. The tool is not installed, or npm’s global folder is not on the agent’s PATH. Run npm install -g @monid-ai/cli and restart Atomic Agent.
The key is rejected. It is missing, mistyped or revoked. Run monid keys list, then add the key again with monid keys add.
BLOCKED. A Monid budget, run cap or the wallet balance stopped the call. Top up or raise the limit at app.monid.ai if you meant to spend more.
The run completed with an error from the provider. The input did not match what the tool expects. Ask the agent to run monid inspect on the tool and fix the input. Provider errors are not billed.
A small local model gets lost. The flow has several steps and exact JSON inputs. In our tests a 9B local model did not follow it reliably, and Claude Haiku 5.5 through OpenRouter did. Use a larger model for Monid tasks.
FAQ
Is Monid free?
No. There is no subscription: you top up a wallet and pay per call. Discovering tools and checking prices is free.
Does my data stay on my computer?
Atomic Agent, your files and your chats stay on your computer. Monid calls go out to the web to fetch data, and the request text (search terms, a domain, an image URL) goes to Monid and the tool’s provider.
Can I cap what Monid spends?
Yes. Set a workspace budget or a per-run cap in Monid. Calls over the limit are blocked, not charged.
Does it work with a local model?
Yes, the model that drives the skill can be local or a cloud provider you configured. The data still comes from Monid over the internet, and a larger model follows the steps more reliably.
Checked on 17 Nov 2026 against Atomic Agent v0.6.11, Desktop v0.0.5, monid.ai/SKILL.md v0.1.9 and monid.ai/docs.
Get updates by email
One email when a new version comes out.