agentleFS
attentiondotnetattentiondotnet/davidondrej-skills/skills/research-and-web/deepapi/SKILL.md

deepapi

Skill11 starsChanged 4 months agoMITView on GitHub

Use DeepAPI for scraping and safe email with DEEPAPI_API_BASE_URL and DEEPAPI_API_KEY.

  • Reads credentials

We split this skill into the rules it gives an agent

46 rules. About 0 are common, 12 are distinctive, and 34 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈12

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 34

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Use this skill when the user asks you to scrape public web data or draft/read/send email through DeepAPI.

    line 9
  2. Distinctive:

    If a request fails unexpectedly, fetch https://deepapi.co/deepapi-skill/VERSION.txt — public, no auth, a few bytes.

    line 14
  3. Very specific:

    If it differs from this file's version, fetch https://deepapi.co/deepapi-skill/SKILL.md and overwrite this file with it.

    line 15
  4. Very specific:

    Every API response includes skillVersion.

    line 16
  5. Distinctive:

    Only ever fetch this skill from https://deepapi.co.

    line 17
  6. Very specific:

    Read DEEPAPI_API_BASE_URL from the environment.

    line 21
  7. Distinctive:

    If either value is missing, stop and ask the user for setup.

    line 23
  8. Very specific:

    Never commit, print, log, paste, or expose DEEPAPI_API_KEY.

    line 24
  9. Very specific:

    Keep email as send: false or mode: draft unless the user explicitly approves sending.

    line 32
  10. Very specific:

    Do not pass inbox IDs. Use emailIdentityId or omit it.

    line 33
  11. Distinctive:

    Run the request with the required headers.

    line 39
  12. Very specific:

    If the response has status: running, wait next.afterSecs and call next.method + next.path until status is succeeded or failed.

    line 40
  13. Very specific:

    If error.retryable is true, wait error.retryAfterSecs before retrying.

    line 41
  14. Very specific:

    If the response is HTTP 402 with error.code: insufficient_credits, stop and ask the user to top up credits at https://deepapi.co/credits.

    line 42
  15. Very specific:

    Use POST /v1/scrape/website.

    line 78
  16. Distinctive:

    Start with small result caps such as maxItems or capability-specific limits.

    line 87
  17. Very specific:

    Use POST /v1/scrape/linkedin/profile.

    line 104
  18. Very specific:

    Use POST /v1/scrape/github/profile.

    line 129
  19. Very specific:

    Use POST /v1/scrape/twitter/search.

    line 154
  20. Very specific:

    Use POST /v1/scrape/linkedin/jobs.

    line 181
  21. Very specific:

    Use POST /v1/scrape/linkedin/company.

    line 206
  22. Very specific:

    Use POST /v1/scrape/linkedin/people.

    line 231
  23. Very specific:

    Use POST /v1/scrape/linkedin/posts.

    line 260
  24. Very specific:

    Use POST /v1/scrape/twitter/user.

    line 286
  25. Very specific:

    Use POST /v1/scrape/twitter/replies.

    line 311
  26. Very specific:

    Use POST /v1/scrape/youtube/transcript.

    line 335
  27. Very specific:

    Use POST /v1/scrape/youtube/channel.

    line 358
  28. Very specific:

    Use POST /v1/scrape/youtube/search.

    line 384
  29. Very specific:

    Use POST /v1/scrape/linkedin.

    line 409
  30. Very specific:

    Use POST /v1/scrape/github.

    line 434
  31. Very specific:

    Use POST /v1/scrape/twitter.

    line 459
  32. Very specific:

    Use POST /v1/email/send. Create an email draft from a workspace email identity.

    line 486
  33. Distinctive:

    Keep send=false or mode=draft unless the user explicitly approves sending.

    line 494
  34. Distinctive:

    Do not pass inboxId or inbox_id.

    line 495
  35. Very specific:

    Use GET /v1/email/messages.

    line 510
  36. Very specific:

    Use GET /v1/email/drafts. List pending email drafts for a workspace email identity.

    line 521
  37. Very specific:

    Use POST /v1/email/drafts/{draftId}/send.

    line 532
  38. Distinctive:

    Only send a draft after the user explicitly approves that draft.

    line 540
  39. Very specific:

    Use POST /v1/research/deep.

    line 551
  40. Distinctive:

    Use query for the research question and context only for relevant background.

    line 559
  41. Very specific:

    Use POST /v1/generate/image.

    line 574
  42. Very specific:

    Use POST /v1/search/web. Search the web and return ranked results with title, url, and snippet.

    line 596
  43. Distinctive:

    Use query for the search terms only.

    line 604
  44. Distinctive:

    Treat snippets as page summaries.

    line 606
  45. Very specific:

    Use GET /v1/requests/{requestId}.

    line 619
  46. Distinctive:

    Only poll request ids created by the same API key.

    line 626

≈ “Common” is an estimate: at least 1 in 5 of the 308,232 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

DeepAPI

Use this skill when the user asks you to scrape public web data or draft/read/send email through DeepAPI.

Staying Up To Date

Required Environment

  • Read DEEPAPI_API_BASE_URL from the environment.
  • Read DEEPAPI_API_KEY from the environment.
  • If either value is missing, stop and ask the user for setup.
  • Never commit, print, log, paste, or expose DEEPAPI_API_KEY.

Request Rules

  • Send Authorization: Bearer $DEEPAPI_API_KEY on every request.
  • Send Content-Type: application/json when sending JSON.
  • Send a unique Idempotency-Key for every POST.
  • For scrape work, set explicit maxCostUsd or maxCostMicrousd.
  • Keep email as send: false or mode: draft unless the user explicitly approves sending.
  • Do not pass inbox IDs. Use emailIdentityId or omit it.

Execution Loop

  1. Choose the narrowest endpoint that matches the task.
  2. Build the request from the endpoint schema and examples below.
  3. Run the request with the required headers.
  4. If the response has status: running, wait next.afterSecs and call next.method + next.path until status is succeeded or failed.
  5. If error.retryable is true, wait error.retryAfterSecs before retrying.
  6. If the response is HTTP 402 with error.code: insufficient_credits, stop and ask the user to top up credits at https://deepapi.co/credits. After top-up, retry with the same Idempotency-Key.
  7. Report requestId, status, debitMicrousd, costFinal, and the useful part of output.

Endpoints

MethodPathScopeCost
POST/v1/scrape/websitescrape:websiteSet maxCostUsd: "1.00" unless the user gives a different cap. The route requires maxCostUsd or maxCostMicrousd as the customer spend cap. The final debit is capped by that amount and reported as debitMicrousd.
POST/v1/scrape/linkedin/profilescrape:linkedinSet maxCostUsd: "0.05" unless the user gives a different cap. The route requires maxCostUsd or maxCostMicrousd as the customer spend cap. The final debit is capped by that amount and reported as debitMicrousd.
POST/v1/scrape/github/profilescrape:githubSet maxCostUsd: "0.03" unless the user gives a different cap. The route requires maxCostUsd or maxCostMicrousd as the customer spend cap. The final debit is capped by that amount and reported as debitMicrousd.
POST/v1/scrape/twitter/searchscrape:twitterSet maxCostUsd: "0.03" unless the user gives a different cap. The route requires maxCostUsd or maxCostMicrousd as the customer spend cap. The final debit is capped by that amount and reported as debitMicrousd.
POST/v1/scrape/linkedin/jobsscrape:linkedinSet maxCostUsd: "0.05" unless the user gives a different cap. The route requires maxCostUsd or maxCostMicrousd as the customer spend cap. The final debit is capped by that amount and reported as debitMicrousd.
POST/v1/scrape/linkedin/companyscrape:linkedinSet maxCostUsd: "0.05" unless the user gives a different cap. The route requires maxCostUsd or maxCostMicrousd as the customer spend cap. The final debit is capped by that amount and reported as debitMicrousd.
POST/v1/scrape/linkedin/peoplescrape:linkedinSet maxCostUsd: "0.50" unless the user gives a different cap. The route requires maxCostUsd or maxCostMicrousd as the customer spend cap. The final debit is capped by that amount and reported as debitMicrousd.
POST/v1/scrape/linkedin/postsscrape:linkedinSet maxCostUsd: "0.05" unless the user gives a different cap. The route requires maxCostUsd or maxCostMicrousd as the customer spend cap. The final debit is capped by that amount and reported as debitMicrousd.
POST/v1/scrape/twitter/userscrape:twitterSet maxCostUsd: "0.05" unless the user gives a different cap. The route requires maxCostUsd or maxCostMicrousd as the customer spend cap. The final debit is capped by that amount and reported as debitMicrousd.
POST/v1/scrape/twitter/repliesscrape:twitterSet maxCostUsd: "0.20" unless the user gives a different cap. The route requires maxCostUsd or maxCostMicrousd as the customer spend cap. The final debit is capped by that amount and reported as debitMicrousd.
POST/v1/scrape/youtube/transcriptscrape:youtubeSet maxCostUsd: "0.05" unless the user gives a different cap. The route requires maxCostUsd or maxCostMicrousd as the customer spend cap. The final debit is capped by that amount and reported as debitMicrousd.
POST/v1/scrape/youtube/channelscrape:youtubeSet maxCostUsd: "0.30" unless the user gives a different cap. The route requires maxCostUsd or maxCostMicrousd as the customer spend cap. The final debit is capped by that amount and reported as debitMicrousd.
POST/v1/scrape/youtube/searchscrape:youtubeSet maxCostUsd: "0.10" unless the user gives a different cap. The route requires maxCostUsd or maxCostMicrousd as the customer spend cap. The final debit is capped by that amount and reported as debitMicrousd.
POST/v1/scrape/linkedinscrape:linkedinSet maxCostUsd: "0.05" unless the user gives a different cap. The route requires maxCostUsd or maxCostMicrousd as the customer spend cap. The final debit is capped by that amount and reported as debitMicrousd.
POST/v1/scrape/githubscrape:githubSet maxCostUsd: "0.03" unless the user gives a different cap. The route requires maxCostUsd or maxCostMicrousd as the customer spend cap. The final debit is capped by that amount and reported as debitMicrousd.
POST/v1/scrape/twitterscrape:twitterSet maxCostUsd: "0.03" unless the user gives a different cap. The route requires maxCostUsd or maxCostMicrousd as the customer spend cap. The final debit is capped by that amount and reported as debitMicrousd.
POST/v1/email/sendemail:sendUses configured email unit pricing; the route does not accept maxCostUsd. Check debitMicrousd in the response.
GET/v1/email/messagesemail:readRead route returns debitMicrousd 0.
GET/v1/email/draftsemail:readRead route returns debitMicrousd 0.
POST/v1/email/drafts/{draftId}/sendemail:sendUses configured email unit pricing; the route does not accept maxCostUsd. Check debitMicrousd in the response.
POST/v1/research/deepresearch:deepSet maxCostUsd: "0.10" unless the user gives a different cap. Defaults to maxCostUsd 0.10. Pass maxCostUsd or maxCostMicrousd to choose a different customer spend cap. The final debit is capped and reported as debitMicrousd.
POST/v1/generate/imagegenerate:imageSet maxCostUsd: "0.20" unless the user gives a different cap. Defaults to maxCostUsd 0.20. Pass maxCostUsd or maxCostMicrousd to choose a different customer spend cap. The final debit is capped and reported as debitMicrousd.
POST/v1/search/websearch:webSet maxCostUsd: "0.05" unless the user gives a different cap. Defaults to maxCostUsd 0.05. Pass maxCostUsd or maxCostMicrousd to choose a different customer spend cap. The final debit is capped and reported as debitMicrousd.
GET/v1/requests/{requestId}same keyStatus polling does not create a new debit.

Endpoint Details

Scrape Website

Use POST /v1/scrape/website. Crawl website pages and return clean text and markdown per page.

Side effects: Starts a scrape run and may debit credits when the run finishes. Polling: If status is running, wait next.afterSecs and call next.method next.path until status is succeeded or failed.

Safety:

  • Send Authorization: Bearer $DEEPAPI_API_KEY and never expose the key.
  • Send a unique Idempotency-Key for every POST.
  • Set an explicit customer spend cap with maxCostUsd or maxCostMicrousd before starting a scrape.
  • Start with small result caps such as maxItems or capability-specific limits.
  • Poll next.path while status is running and report the final debitMicrousd.

Example body:

{
  "maxCostUsd": "1.00",
  "waitForFinishSecs": 60,
  "urls": [
    "https://example.com"
  ],
  "maxPages": 1
}

Scrape LinkedIn Profile

Use POST /v1/scrape/linkedin/profile. Scrape public LinkedIn profile details.

Side effects: Starts a scrape run and may debit credits when the run finishes. Polling: If status is running, wait next.afterSecs and call next.method next.path until status is succeeded or failed.

Safety:

  • Send Authorization: Bearer $DEEPAPI_API_KEY and never expose the key.
  • Send a unique Idempotency-Key for every POST.
  • Set an explicit customer spend cap with maxCostUsd or maxCostMicrousd before starting a scrape.
  • Start with small result caps such as maxItems or capability-specific limits.
  • Poll next.path while status is running and report the final debitMicrousd.

Example body:

{
  "maxCostUsd": "0.05",
  "waitForFinishSecs": 60,
  "profiles": [
    "williamhgates"
  ]
}

Scrape GitHub Profile

Use POST /v1/scrape/github/profile. Scrape public GitHub profile details.

Side effects: Starts a scrape run and may debit credits when the run finishes. Polling: If status is running, wait next.afterSecs and call next.method next.path until status is succeeded or failed.

Safety:

  • Send Authorization: Bearer $DEEPAPI_API_KEY and never expose the key.
  • Send a unique Idempotency-Key for every POST.
  • Set an explicit customer spend cap with maxCostUsd or maxCostMicrousd before starting a scrape.
  • Start with small result caps such as maxItems or capability-specific limits.
  • Poll next.path while status is running and report the final debitMicrousd.

Example body:

{
  "maxCostUsd": "0.03",
  "waitForFinishSecs": 60,
  "usernames": [
    "octocat"
  ]
}

Search X/Twitter

Use POST /v1/scrape/twitter/search. Scrape X/Twitter posts from a search query or account handles.

Side effects: Starts a scrape run and may debit credits when the run finishes. Polling: If status is running, wait next.afterSecs and call next.method next.path until status is succeeded or failed.

Safety:

  • Send Authorization: Bearer $DEEPAPI_API_KEY and never expose the key.
  • Send a unique Idempotency-Key for every POST.
  • Set an explicit customer spend cap with maxCostUsd or maxCostMicrousd before starting a scrape.
  • Start with small result caps such as maxItems or capability-specific limits.
  • Poll next.path while status is running and report the final debitMicrousd.

Example body:

{
  "maxCostUsd": "0.03",
  "waitForFinishSecs": 60,
  "handles": [
    "nasa"
  ],
  "maxItems": 1,
  "sort": "latest"
}

Scrape LinkedIn Jobs

Use POST /v1/scrape/linkedin/jobs. Scrape public LinkedIn job listings for a search query.

Side effects: Starts a scrape run and may debit credits when the run finishes. Polling: If status is running, wait next.afterSecs and call next.method next.path until status is succeeded or failed.

Safety:

  • Send Authorization: Bearer $DEEPAPI_API_KEY and never expose the key.
  • Send a unique Idempotency-Key for every POST.
  • Set an explicit customer spend cap with maxCostUsd or maxCostMicrousd before starting a scrape.
  • Start with small result caps such as maxItems or capability-specific limits.
  • Poll next.path while status is running and report the final debitMicrousd.

Example body:

{
  "maxCostUsd": "0.05",
  "waitForFinishSecs": 60,
  "query": "software engineer",
  "location": "United States",
  "maxItems": 5
}

Scrape LinkedIn Company

Use POST /v1/scrape/linkedin/company. Scrape public LinkedIn company pages for firmographic details.

Side effects: Starts a scrape run and may debit credits when the run finishes. Polling: If status is running, wait next.afterSecs and call next.method next.path until status is succeeded or failed.

Safety:

  • Send Authorization: Bearer $DEEPAPI_API_KEY and never expose the key.
  • Send a unique Idempotency-Key for every POST.
  • Set an explicit customer spend cap with maxCostUsd or maxCostMicrousd before starting a scrape.
  • Start with small result caps such as maxItems or capability-specific limits.
  • Poll next.path while status is running and report the final debitMicrousd.

Example body:

{
  "maxCostUsd": "0.05",
  "waitForFinishSecs": 60,
  "companies": [
    "microsoft"
  ]
}

Search LinkedIn People

Use POST /v1/scrape/linkedin/people. Search public LinkedIn profiles by role, location, company, or school. Requires maxCostUsd of at least 0.50.

Side effects: Starts a scrape run and may debit credits when the run finishes. Polling: If status is running, wait next.afterSecs and call next.method next.path until status is succeeded or failed.

Safety:

  • Send Authorization: Bearer $DEEPAPI_API_KEY and never expose the key.
  • Send a unique Idempotency-Key for every POST.
  • Set an explicit customer spend cap with maxCostUsd or maxCostMicrousd before starting a scrape.
  • Start with small result caps such as maxItems or capability-specific limits.
  • Poll next.path while status is running and report the final debitMicrousd.

Example body:

{
  "maxCostUsd": "0.50",
  "waitForFinishSecs": 60,
  "titles": [
    "Founder"
  ],
  "locations": [
    "San Francisco"
  ],
  "maxItems": 5
}

Scrape LinkedIn Posts

Use POST /v1/scrape/linkedin/posts. Scrape recent public posts from LinkedIn profiles or company pages.

Side effects: Starts a scrape run and may debit credits when the run finishes. Polling: If status is running, wait next.afterSecs and call next.method next.path until status is succeeded or failed.

Safety:

  • Send Authorization: Bearer $DEEPAPI_API_KEY and never expose the key.
  • Send a unique Idempotency-Key for every POST.
  • Set an explicit customer spend cap with maxCostUsd or maxCostMicrousd before starting a scrape.
  • Start with small result caps such as maxItems or capability-specific limits.
  • Poll next.path while status is running and report the final debitMicrousd.

Example body:

{
  "maxCostUsd": "0.05",
  "waitForFinishSecs": 60,
  "profiles": [
    "williamhgates"
  ],
  "maxItems": 3
}

Scrape X/Twitter User

Use POST /v1/scrape/twitter/user. Scrape public X/Twitter account profiles, with optional follower and following lists.

Side effects: Starts a scrape run and may debit credits when the run finishes. Polling: If status is running, wait next.afterSecs and call next.method next.path until status is succeeded or failed.

Safety:

  • Send Authorization: Bearer $DEEPAPI_API_KEY and never expose the key.
  • Send a unique Idempotency-Key for every POST.
  • Set an explicit customer spend cap with maxCostUsd or maxCostMicrousd before starting a scrape.
  • Start with small result caps such as maxItems or capability-specific limits.
  • Poll next.path while status is running and report the final debitMicrousd.

Example body:

{
  "maxCostUsd": "0.05",
  "waitForFinishSecs": 60,
  "handles": [
    "nasa"
  ]
}

Scrape X/Twitter Replies

Use POST /v1/scrape/twitter/replies. Scrape the public reply thread of an X/Twitter post. Requires maxCostUsd of at least 0.20.

Side effects: Starts a scrape run and may debit credits when the run finishes. Polling: If status is running, wait next.afterSecs and call next.method next.path until status is succeeded or failed.

Safety:

  • Send Authorization: Bearer $DEEPAPI_API_KEY and never expose the key.
  • Send a unique Idempotency-Key for every POST.
  • Set an explicit customer spend cap with maxCostUsd or maxCostMicrousd before starting a scrape.
  • Start with small result caps such as maxItems or capability-specific limits.
  • Poll next.path while status is running and report the final debitMicrousd.

Example body:

{
  "maxCostUsd": "0.20",
  "waitForFinishSecs": 60,
  "url": "https://x.com/NASA/status/1234567890123456789",
  "maxItems": 5
}

Scrape YouTube Transcript

Use POST /v1/scrape/youtube/transcript. Scrape the transcript of a YouTube video as plain text plus timed segments. Videos without captions return an empty result.

Side effects: Starts a scrape run and may debit credits when the run finishes. Polling: If status is running, wait next.afterSecs and call next.method next.path until status is succeeded or failed.

Safety:

  • Send Authorization: Bearer $DEEPAPI_API_KEY and never expose the key.
  • Send a unique Idempotency-Key for every POST.
  • Set an explicit customer spend cap with maxCostUsd or maxCostMicrousd before starting a scrape.
  • Start with small result caps such as maxItems or capability-specific limits.
  • Poll next.path while status is running and report the final debitMicrousd.

Example body:

{
  "maxCostUsd": "0.05",
  "waitForFinishSecs": 60,
  "url": "https://www.youtube.com/watch?v=dQw4w9WgXcQ"
}

Scrape YouTube Channel

Use POST /v1/scrape/youtube/channel. Scrape a YouTube channel's stats and recent videos. Each video item includes subscriber and channel totals.

Side effects: Starts a scrape run and may debit credits when the run finishes. Polling: If status is running, wait next.afterSecs and call next.method next.path until status is succeeded or failed.

Safety:

  • Send Authorization: Bearer $DEEPAPI_API_KEY and never expose the key.
  • Send a unique Idempotency-Key for every POST.
  • Set an explicit customer spend cap with maxCostUsd or maxCostMicrousd before starting a scrape.
  • Start with small result caps such as maxItems or capability-specific limits.
  • Poll next.path while status is running and report the final debitMicrousd.

Example body:

{
  "maxCostUsd": "0.30",
  "waitForFinishSecs": 60,
  "channels": [
    "mkbhd"
  ],
  "maxItems": 3
}

Search YouTube

Use POST /v1/scrape/youtube/search. Search YouTube videos by keyword and return video metadata.

Side effects: Starts a scrape run and may debit credits when the run finishes. Polling: If status is running, wait next.afterSecs and call next.method next.path until status is succeeded or failed.

Safety:

  • Send Authorization: Bearer $DEEPAPI_API_KEY and never expose the key.
  • Send a unique Idempotency-Key for every POST.
  • Set an explicit customer spend cap with maxCostUsd or maxCostMicrousd before starting a scrape.
  • Start with small result caps such as maxItems or capability-specific limits.
  • Poll next.path while status is running and report the final debitMicrousd.

Example body:

{
  "maxCostUsd": "0.10",
  "waitForFinishSecs": 60,
  "query": "ai agents",
  "sort": "views",
  "maxItems": 3
}

Scrape LinkedIn

Use POST /v1/scrape/linkedin. Backward-compatible alias for LinkedIn profile scraping.

Side effects: Starts a scrape run and may debit credits when the run finishes. Polling: If status is running, wait next.afterSecs and call next.method next.path until status is succeeded or failed.

Safety:

  • Send Authorization: Bearer $DEEPAPI_API_KEY and never expose the key.
  • Send a unique Idempotency-Key for every POST.
  • Set an explicit customer spend cap with maxCostUsd or maxCostMicrousd before starting a scrape.
  • Start with small result caps such as maxItems or capability-specific limits.
  • Poll next.path while status is running and report the final debitMicrousd.

Example body:

{
  "maxCostUsd": "0.05",
  "waitForFinishSecs": 60,
  "profiles": [
    "williamhgates"
  ]
}

Scrape GitHub

Use POST /v1/scrape/github. Backward-compatible alias for GitHub profile scraping.

Side effects: Starts a scrape run and may debit credits when the run finishes. Polling: If status is running, wait next.afterSecs and call next.method next.path until status is succeeded or failed.

Safety:

  • Send Authorization: Bearer $DEEPAPI_API_KEY and never expose the key.
  • Send a unique Idempotency-Key for every POST.
  • Set an explicit customer spend cap with maxCostUsd or maxCostMicrousd before starting a scrape.
  • Start with small result caps such as maxItems or capability-specific limits.
  • Poll next.path while status is running and report the final debitMicrousd.

Example body:

{
  "maxCostUsd": "0.03",
  "waitForFinishSecs": 60,
  "usernames": [
    "octocat"
  ]
}

Scrape Twitter

Use POST /v1/scrape/twitter. Backward-compatible alias for X/Twitter search scraping.

Side effects: Starts a scrape run and may debit credits when the run finishes. Polling: If status is running, wait next.afterSecs and call next.method next.path until status is succeeded or failed.

Safety:

  • Send Authorization: Bearer $DEEPAPI_API_KEY and never expose the key.
  • Send a unique Idempotency-Key for every POST.
  • Set an explicit customer spend cap with maxCostUsd or maxCostMicrousd before starting a scrape.
  • Start with small result caps such as maxItems or capability-specific limits.
  • Poll next.path while status is running and report the final debitMicrousd.

Example body:

{
  "maxCostUsd": "0.03",
  "waitForFinishSecs": 60,
  "handles": [
    "nasa"
  ],
  "maxItems": 1,
  "sort": "latest"
}

Send Email

Use POST /v1/email/send. Create an email draft from a workspace email identity; set send=true to send it.

Side effects: Creates a draft, or sends an email when direct send is approved. Polling: This route returns a terminal envelope directly.

Safety:

  • Send Authorization: Bearer $DEEPAPI_API_KEY and never expose the key.
  • Send a unique Idempotency-Key for every POST.
  • Keep send=false or mode=draft unless the user explicitly approves sending.
  • Do not pass inboxId or inbox_id; use emailIdentityId or the workspace default.
  • Attachments, hidden HTML, image HTML, URL shorteners, and high-risk direct sends are blocked by policy.

Example body:

{
  "to": "<email-address>",
  "subject": "Quick hello",
  "text": "Hi, this is a draft from my agent.",
  "send": false
}

Receive Email

Use GET /v1/email/messages. Read messages for a workspace email identity.

Side effects: Reads messages only. Polling: This route returns a terminal envelope directly.

Safety:

  • Send Authorization: Bearer $DEEPAPI_API_KEY and never expose the key.
  • Do not pass inboxId or inbox_id; use emailIdentityId or the workspace default.

List Drafts

Use GET /v1/email/drafts. List pending email drafts for a workspace email identity.

Side effects: Reads drafts only. Polling: This route returns a terminal envelope directly.

Safety:

  • Send Authorization: Bearer $DEEPAPI_API_KEY and never expose the key.
  • Do not pass inboxId or inbox_id; use emailIdentityId or the workspace default.

Send Draft

Use POST /v1/email/drafts/{draftId}/send. Approve and send an existing draft by draftId after review.

Side effects: Sends the reviewed draft as a real email when direct send is approved. Polling: This route returns a terminal envelope directly.

Safety:

  • Send Authorization: Bearer $DEEPAPI_API_KEY and never expose the key.
  • Send a unique Idempotency-Key for every POST.
  • Only send a draft after the user explicitly approves that draft.
  • Do not pass inboxId or inbox_id; use emailIdentityId or the workspace default.
  • Sending re-checks recipient and content policy against the stored draft; blocked drafts stay drafts.

Example body:

{}

Deep Research

Use POST /v1/research/deep. Answer a research question with current web evidence.

Side effects: Runs a paid web research request and debits credits when finished. Polling: This route returns a terminal envelope directly.

Safety:

  • Send Authorization: Bearer $DEEPAPI_API_KEY and never expose the key.
  • Send a unique Idempotency-Key for every POST.
  • Use query for the research question and context only for relevant background.
  • Set maxCostUsd when you need a lower or higher spend cap than the default.
  • Report debitMicrousd and summarize the returned sources when sources are present.

Example body:

{
  "query": "What changed in EU AI Act compliance timelines for API startups?",
  "context": "We sell API tooling to EU customers.",
  "maxCostUsd": "0.10"
}

Generate Image

Use POST /v1/generate/image. Generate an image from a text prompt.

Side effects: Runs a paid image generation request and debits credits when finished. Polling: This route returns a terminal envelope directly.

Safety:

  • Send Authorization: Bearer $DEEPAPI_API_KEY and never expose the key.
  • Send a unique Idempotency-Key for every POST.
  • Describe the image you want in prompt, including style and composition.
  • Set maxCostUsd when you need a lower or higher spend cap than the default.
  • output.images contains base64 data URLs; save them to files instead of printing them.

Example body:

{
  "prompt": "A minimal flat illustration of a rocket launching from a laptop screen",
  "maxCostUsd": "0.20"
}

Web Search

Use POST /v1/search/web. Search the web and return ranked results with title, url, and snippet.

Side effects: Runs a paid web search request and debits credits when finished. Polling: This route returns a terminal envelope directly.

Safety:

  • Send Authorization: Bearer $DEEPAPI_API_KEY and never expose the key.
  • Send a unique Idempotency-Key for every POST.
  • Use query for the search terms only; keep it under 500 characters.
  • Set maxCostUsd when you need a lower or higher spend cap than the default.
  • Treat snippets as page summaries; open a result URL when you need the full content.

Example body:

{
  "query": "latest stable Node.js LTS version",
  "maxResults": 3,
  "maxCostUsd": "0.05"
}

Request Status

Use GET /v1/requests/{requestId}. Poll a running request by requestId.

Side effects: Reads or refreshes request status. Polling: If status is running, wait next.afterSecs and call next.method next.path until status is succeeded or failed.

Safety:

  • Send Authorization: Bearer $DEEPAPI_API_KEY and never expose the key.
  • Only poll request ids created by the same API key.

Example query: waitForFinishSecs=60

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.