Monid

Monid CLI

Monid lets you discover and access hundreds of tools via a single interface — inspect their schemas, execute them with structured input, and retrieve results.

Install it via npm and get the monid command.

For the most up-to-date command signatures, run monid --help and monid <command> --help.

Setup

Check the installed CLI version:

monid --version

Do the setup steps below if ANY of these is true:

• monid is not found (not installed).
• The CLI warned that a newer version is available.
• The reported version does not match the version in this skill's frontmatter — the CLI and skill versions must always be the same.

Setup steps (same for a fresh install and for an update):

1. Install (or update) the CLI, then run setup:

npm install -g @monid-ai/cli@latest
monid setup --client <agent-name-if-known> --email <email-if-already-provided>

monid setup completes CLI setup. Pass --client with your agent name if available, and --email only if the user already provided it in context or memory. Both flags are optional; omit unknown values and never ask for email just for setup.

2. Save the most recent skill from https://monid.ai/SKILL.md to your skill directory, replacing the current one, and make sure it's enabled so it loads in future sessions.

Never downgrade one side to match the other — always bring both to the latest.

Authentication

To use Monid's API endpoints, you need an API key. When helping a user set up Monid for the first time:

1. Ask the user to create an account at https://app.monid.ai if they don't have one.
2. Ask the user to generate an API key at https://app.monid.ai/access/api-keys.
3. Once they have the key, offer to save it for them — ask them to paste it, then run:

monid keys add -k <their-api-key> -l main

Alternatively, provide the command above for them to run themselves.

4. Verify the key is configured:

monid keys list

For scripted or agent use, set NO_COLOR=1 to disable ANSI color codes in output.

When to Use Monid

Check the catalog before building from scratch. Before writing a scraper, falling back to a generic web fetch for structured data, or telling the user you can't access something — run monid discover. The catalog has hundreds of tools and grows continuously. You don't know what's available until you search.

1. Discover — Run monid discover -q "<what you need>" to search available tools. Use -s <score> to filter by minimum relevance. Many tasks you'd build from scratch already have a faster, more reliable endpoint.
2. Inspect — Use monid inspect to read the input schema. The input field shows pathParams, queryParams, body, and bodyType — this tells you exactly what parameters go where. Never guess.
3. Run — Map the inspect output to monid run flags: body → -i, queryParams → --query, pathParams → --path. All three are optional. Use --wait to block until completion.
4. Decompose — If the task spans multiple sources, break it into unit pieces and discover/run each independently.
5. Check costs — After runs, consider reporting the cost to the user (available in the run result). Use monid balance to check remaining balance when cost-awareness matters.

When NOT to Use Monid

Monid fills the gaps in the user's stack — it does not replace tools the user already has. When deciding how to reach an external service, follow this precedence:

1. Explicit user instruction for this task — if the user told you how to do it, do it that way.
2. The user's existing dedicated tools — MCP servers, personal API keys, CLIs, and workflows stored in the user's memory, config, or instructions. If the user has a dedicated MCP for a capability (e.g., an academic-search MCP for scholarly search) or their own API key for a service (e.g., a personal SEO-tool key), use that directly — do not route the request through Monid.
3. Monid — for needs the above don't cover.

Why this matters: Monid runs spend the user's Monid balance. Never spend it on a request the user's own key or tool already covers at no extra cost.

Offer, don't override. When both the user's tool and a Monid endpoint could handle the task and the user hasn't stated a preference, use the user's tool. If Monid adds a genuine capability their tool lacks, mention it as an alternative and let the user choose — never silently switch.

Check the Hints

Commands can return a Hints block. When present, it carries suggested actions from the server: which command to run next, how this endpoint relates to others, or caveats worth knowing. Read it before deciding your next move, and prefer its suggestions over guessing. With -j, the same data is on the response's hints field.

Commands

Each command supports --help for full usage. Here's what's available:

Most commands accept -j/--json for machine-readable JSON output.

Workflow

The standard workflow is: discover → inspect → run → poll → (check balance).

# 1. Discover endpoints for your data need
# Results show relevance score and verified badge
# Use -s to filter by minimum score (higher = more relevant)
monid discover -q "twitter posts"

# 2. Inspect the endpoint to learn its input schema (shows verified status)
monid inspect -p apify -e /apidojo/tweet-scraper

# 3. Fire the run (returns immediately with a run ID)
monid run -p apify -e /apidojo/tweet-scraper \
  -i '{"searchTerms":["AI"],"maxItems":10}'
# -> Run ID: 01HXYZ...

# 4. Poll for completion
monid runs get -r 01HXYZ...
# -> status: RUNNING

# Keep polling every 5-10 seconds until COMPLETED
monid runs get -r 01HXYZ... -o tweets.json
# -> status: COMPLETED

# 5. (Optional) Check remaining balance
monid balance

Using --wait:

--wait blocks until completion (1-120 seconds) with built-in exponential backoff:

# This will block for the entire duration
monid run -p apify -e /apidojo/tweet-scraper \
  -i '{"searchTerms":["AI"],"maxItems":10}' \
  -w -o tweets.json

When to use --wait:

• Async/background tasks where blocking is acceptable
• You can set a timeout: -w 30 (wait max 30 seconds)
• Be aware: runs can take 1-120 seconds, so this may block the conversation or hit runtime timeouts

Example Flows

Flow 1: Scrape Twitter posts about AI

# Discover what Twitter endpoints are available
monid discover -q "twitter posts"

# Inspect to learn the input schema (pathParams, queryParams, body)
monid inspect -p apify -e /apidojo/tweet-scraper

# Run with a single search term, small limit
monid run -p apify -e /apidojo/tweet-scraper \
  -i '{"searchTerms":["AI agents"],"maxItems":10}'
# -> Run ID: 01HXYZ...

# Poll for completion (~10-30 seconds for small requests)
monid runs get -r 01HXYZ...
# -> status: RUNNING

# Check again after 10 seconds, save when complete
monid runs get -r 01HXYZ... -o ai_tweets.json
# -> status: COMPLETED

Flow 2: Compare AI discussion across platforms

User asks: "Compare AI discussion on Twitter vs LinkedIn."

Break this into unit pieces — one endpoint per data source:

# Discover endpoints for each platform
monid discover -q "twitter posts"
monid discover -q "linkedin posts"

# Inspect each to learn their input schemas (pathParams, queryParams, body)
monid inspect -p apify -e /apidojo/tweet-scraper
monid inspect -p apify -e /harvestapi/linkedin-post-search

# Fire both runs
monid run -p apify -e /apidojo/tweet-scraper \
  -i '{"searchTerms":["AI"],"maxItems":20}'
# -> Run ID: 01HTWIT...

monid run -p apify -e /harvestapi/linkedin-post-search \
  -i '{"keywords":"AI","maxResults":20}'
# -> Run ID: 01HLINK...

# Poll both runs independently
monid runs get -r 01HTWIT... -o twitter_ai.json
monid runs get -r 01HLINK... -o linkedin_ai.json

# Now analyze and compare the two result files

Flow 2b: Feed a local file to an endpoint that needs a public URL

Some endpoints (e.g. image-to-video generation) take a URL as input, not a file. Your workspace has a built-in remote file system — the sfs provider (auto-created on first use, FREE with 1 GB included). Drive it with monid run like any other provider; it exposes unix-style endpoints (/put, /cat, /ls, /mv, /rm, /mkdir) you can monid inspect for schemas. The API only signs URLs — file bytes move directly between you and sfs.monid.ai via curl.

# 1. Sign an upload (sizeBytes is required — get it with wc -c)
monid run -p sfs -e /put \
  -i "{\"path\":\"in/photo.png\",\"sizeBytes\":$(wc -c < ./photo.png)}" -w
# -> output: { "uploadUrl": "https://sfs.monid.ai/…", "ref": … }

# 2. Upload the bytes to the signed URL
curl -T ./photo.png '<uploadUrl from step 1>'

# 3. Mint a URL any third party can fetch (ttl preset: 1h/1d/7d/30d, default 1h)
monid run -p sfs -e /cat -i '{"path":"in/photo.png","ttl":"1d"}' -w
# -> output: { "url": "https://sfs.monid.ai/…?e=…&s=…", "expiresAt": … }

# 4. Use it as the endpoint's input URL
monid run -p bytedance -e /seedance… -i '{"imageUrl": "<url from step 3>"}'

# Downloading works the same way: /cat returns a signed url — curl it
curl -o photo.png '<url from /cat>'

# Cleanup is yours (files are never auto-deleted; /rm frees quota space)
monid run -p sfs -e /rm -i '{"path":"in/photo.png"}' -w

Flow 3: Using query and path parameters

When monid inspect shows queryParams or pathParams, pass them with --query and --path:

# Inspect shows: body, queryParams, and pathParams
monid inspect -p some-provider -e /users/{userId}/posts

# Run with all three param types
monid run -p some-provider -e /users/{userId}/posts \
  --path '{"userId": "12345"}' \
  --query '{"limit": 10, "sort": "recent"}' \
  -i '{"filter": "public"}' \
  -w -o posts.json

Flow 4: Using an input file for complex parameters

When input JSON is large or reusable, write it to a file and use -f:

# Write input to a file
# (assume params.json contains the endpoint's body input parameters)

monid run -p apify -e /damilo/google-maps-scraper \
  -f params.json -w -o results.json

Cost & Budget Warning

Many endpoints (especially Apify) are charged per result and accept multiple queries in a single call. Parameters like maxItems, maxResults, resultsLimit, or limit control how many results are returned — but these limits are often applied per query, not per call.

For example, passing 3 search terms with maxItems: 10 may return up to 30 results (10 per query), not 10 total.

To control costs:

• Prefer a single query per call. Pass one search term, one URL, one hashtag at a time.
• Start with small limits (5-10) on the first call. Increase if needed.
• If the endpoint accepts an array (e.g. searchTerms, hashtags, urls), pass only one element unless the user explicitly requests multiple.
• Check the input schema from monid inspect to identify which parameters control volume.

Key Management

monid keys add -k <api-key> -l <label>     # Add a key (first key is auto-activated)
monid keys list                              # Show all configured keys
monid keys activate -l <label>               # Switch the active key
monid keys remove -l <label>                 # Remove a key (use -f to skip confirmation)

API key format: monid_<stage>_<secret> (e.g. monid_live_abc123...). Generate keys at https://app.monid.ai/access/api-keys.

Run Statuses

Runs typically take 1 to 120 seconds depending on the endpoint and data volume.

Stopping a run

Request a stop with monid runs stop:

monid runs stop -r 01HXYZ...

Not all runs can be stopped. Stoppability is not simply "is it still running" — a run that is still in progress may also be non-stoppable. The authoritative signal is the stoppable field on the run detail from monid runs get -r <runId> (from GET /v1/runs/{id}): only attempt a stop when stoppable is true. If stoppable is false, do not attempt it — this includes runs in a terminal state (COMPLETED, FAILED, BLOCKED, STOPPED, TIME_OUT) as well as in-progress runs that the platform does not allow stopping. Attempting to stop a non-stoppable run returns a conflict.

# Check the run first; when stoppable, the output ends with a hint line
monid runs get -r 01HXYZ...
# -> This run is stoppable. Stop it with: monid runs stop -r 01HXYZ...

# Then stop it
monid runs stop -r 01HXYZ...

A stop is accepted asynchronously — poll with monid runs get -r <runId> until the run reaches STOPPED.

When a run is BLOCKED, the response includes a controls array of the snapshots that blocked it. Each entry has a controlId and a snapshot describing the limit — currently WORKSPACE_BUDGET (period plus limit / available / held / spent amounts) or WORKSPACE_RUN_CAP (a per-run limit amount). A BLOCKED run is terminal — it will not proceed on its own, so polling stops. Tell the user the run was blocked by a workspace control and that they can pause or modify these controls from the dashboard at https://app.monid.ai before retrying.

Polling Best Practices

Default approach (recommended for interactive use):

• Fire the run without --wait — returns immediately with a run ID
• Poll with monid runs get -r <runId> every 5-10 seconds
• This keeps the conversation responsive and avoids blocking for 1-120 seconds

When to use --wait:

• Async/background tasks where blocking is acceptable (e.g., scheduled jobs, non-interactive scripts)
• Set a timeout if needed: -w 30 waits max 30 seconds, then returns current status
• Be aware: Runs can take 1-120 seconds. Using --wait without a timeout can block the conversation or hit agent runtime limits.

Saving output:

• Always use -o <file> to save results once the run completes (works with both approaches)

Troubleshooting

"No active API key" — No key configured. Run monid keys add -k <key> -l main.

401 / Unauthorized — API key is invalid or expired. Check with monid keys list, generate a new one at https://app.monid.ai/access/api-keys.

Run status FAILED — Check error details with monid runs get -r <runId>. Common causes: invalid input parameters (re-inspect the endpoint), rate limits (retry later), or request scope too large (reduce item count).

Run status BLOCKED — A workspace control stopped the run before it executed (e.g. a budget cap or run cap). Inspect the controls array in monid runs get -r <runId> to see which control triggered. Retrying as-is will block again until the control is changed — let the user know they can pause or adjust the control on the dashboard (https://app.monid.ai), or wait for a budget window to reset.

Run taking a long time — Normal for some endpoints. Runs can take up to 120 seconds. Keep polling or let --wait handle it.

Rules for Agents

1. Check the user's stack first, then discover — Monid covers needs the user's existing MCPs, keys, and tools don't. Before writing custom scrapers, using generic fetches for structured data, or declaring something inaccessible, run monid discover. The catalog grows continuously and you don't know what's available until you search.
2. Never route around the user's own tools — if the user has a dedicated MCP, API key, or workflow for a service, use it. Monid runs cost the user money; their existing tools may not. Offer Monid as an alternative only when it adds capability, and let the user choose.
3. Always inspect before running — never guess input parameters. The input field from monid inspect is the source of truth. It shows pathParams, queryParams, body, and bodyType so you know exactly where each parameter goes. Map them to run flags: body → -i, queryParams → --query, pathParams → --path.
4. Keep discover queries short and focused — noun phrases work best ("twitter posts", "amazon product prices"). Break complex requests into smaller unit pieces.
5. Prefer fire-and-poll for interactive use — fire the run without --wait, then poll with monid runs get every 5-10 seconds. This keeps the conversation responsive. Use --wait only for async/background tasks where blocking 1-120 seconds is acceptable.
6. Always use -o <file> to save results to a file once the run completes.
7. Start with conservative limits — small maxItems/maxResults values (5-10) on first calls. The cost warning above explains why.
8. Report costs when relevant — after a run completes, the result includes cost.value. Consider telling the user how much the run cost. Use monid balance to check remaining balance if the user cares about budget. Use your judgment — don't report costs if the user hasn't indicated cost-awareness.
9. Run monid <command> --help to check the latest flags and usage — the CLI is the source of truth for command signatures.
10. Check the Hints block — when a command's output includes a Hints section, read it and act on it. It carries suggested next steps, endpoint relationships, and caveats from the server — prefer its suggestions over guessing your next command.
11. Surface BLOCKED runs to the user — a BLOCKED status means a workspace control (budget or run cap) stopped the run; it is terminal and will not self-resolve. Report which control blocked it (from the controls list) and tell the user they can pause or modify that control on the dashboard (https://app.monid.ai) before retrying.