> Parel documentation. Index of every page: https://parel.si/docs/llms.txt. This page in HTML: https://parel.si/docs/changelog

# Changelog

> What changed for customers, newest first. Internal work, operations and founder steps are left out.

## 2026-11-20: domain history charges every lookup

- **[`get_domain_history`](https://parel.si/docs/tools/get_domain_history.md)** now charges a lookup that finds nothing like any other, at what the lookup really cost: the source bills for the lookup, whatever it finds. A failed call is still ₹0, and every call still shows its price and asks first. [`get_domain_overview`](https://parel.si/docs/tools/get_domain_overview.md) is the far cheaper way to see whether a domain has rankings at all.

## 2026-11-20: research at the depth you choose

- **[`research_brief`](https://parel.si/docs/tools/research_brief.md)**: a researched answer to one question, written from many live sources with numbered citations. Two levels, `lite` (seconds) and `standard` (up to a minute), each with its own price.
- **[Research levels](https://parel.si/docs/research.md)**: every way Parel researches a question, from a quick cited answer to a written brief, with what each gives, how long it takes and today's price. A quote is free, so your AI can show you the levels and prices before running one.
- **Three new [playbooks](https://parel.si/playbooks)**: map a market, find the people worth hearing from (podcasts, papers, people, creators), and pick the right spots at the right hour (places and their busy times).
- **[`search_places_by_description`](https://parel.si/docs/tools/search_places_by_description.md)**: places described in plain words ("quiet cafés in Indiranagar good for a laptop"), each with address, rating, opening hours and how busy it is, hour by hour.
- **[`search_fast`](https://parel.si/docs/tools/search_fast.md)**: the fastest, cheapest search by meaning, for agents that run many small searches.
- **[Search by meaning](https://parel.si/docs/search-by-meaning.md)**: one page for every search that takes a description instead of keywords: pages, podcasts, people, companies, creators, places, papers and news. [`search_semantic`](https://parel.si/docs/tools/search_semantic.md) explains its `financial report` and `personal site` filters.

## 2026-11-20: eleven more tools: Google suggestions, the full results page, page speed, Trends by state, competitors, ad, ranking and backlink history, Instagram contacts, Threads replies and hashtag search

- **Eleven tools switched on** after a live call each: [`get_search_suggestions`](https://parel.si/docs/tools/get_search_suggestions.md) (what Google suggests as people type), [`get_search_page`](https://parel.si/docs/tools/get_search_page.md) (everything on Google's first page: People also ask, the AI Overview and the pages it cites, related searches), [`check_page_speed`](https://parel.si/docs/tools/check_page_speed.md) (Lighthouse scores for one page, desktop or mobile), [`get_domain_competitors`](https://parel.si/docs/tools/get_domain_competitors.md) (the sites competing with a domain in Google), [`get_domain_history`](https://parel.si/docs/tools/get_domain_history.md) (a domain's rankings and traffic month by month; it quotes its price and runs only with `confirm: true`), [`get_trends_by_region`](https://parel.si/docs/tools/get_trends_by_region.md) (Google Trends interest state by state), [`get_backlink_changes`](https://parel.si/docs/tools/get_backlink_changes.md) (backlinks gained and lost each month), [`get_ad_history`](https://parel.si/docs/tools/get_ad_history.md) (a domain's past Google search ads, priced per ad), [`get_instagram_contact`](https://parel.si/docs/tools/get_instagram_contact.md) (the email and phone an Instagram account publishes), [`get_threads_comments`](https://parel.si/docs/tools/get_threads_comments.md) (replies under a Threads post) and [`search_instagram_hashtags`](https://parel.si/docs/tools/search_instagram_hashtags.md) (hashtags by keyword with their post counts). A call that fails or finds nothing is ₹0, as everywhere.

## 2026-11-20: connect from more apps, and see the moment it works

- **The Connect page in the dashboard is rebuilt**: pick your app, follow one install link or one command, and a box on the page shows when the app has connected and when its first call has run, with what the call cost, or the reason if it was refused (nothing is charged for a refused call).
- **More apps have their own page**: GitHub Copilot CLI, Gemini CLI, OpenCode, Zed, Kiro, Cline, Qwen Code, Goose and any other agent that can read a line of setup, next to Claude, ChatGPT, Claude Code, Cursor, Codex, VS Code, Windsurf and n8n. Start from the [docs](https://parel.si/docs/index.md).
- **Add to Claude in one click**: a link that opens Claude's add box already filled in; you only sign in.
- **Disconnecting disconnects everywhere**: revoking the key named "Claude connector" under API keys now stops every app you signed in from, not only the last one used. Sign in again from an app to reconnect it.

## 2026-11-18: second sources, and a pointer to them when a call fails

- **63 new tools named `<tool>_backup`**, for example [`find_email_backup`](https://parel.si/docs/tools/find_email_backup.md) and [`get_linkedin_profile_backup`](https://parel.si/docs/tools/get_linkedin_profile_backup.md): the same job as the first-choice tool, from another data provider, at their own price. A failed or empty call is still ₹0.
- **When a call fails at the source, or finds nothing**, and a second source can help, the answer carries [`also_try`](https://parel.si/docs/errors.md): the tool to try, why, its estimated price and how to call it. Nothing more runs until you call it. A second source that reads the same platform is suggested only after a failure.
- **[`discover_tools`](https://parel.si/docs/mcp.md#discover_tools)** lists the second sources when you name one or ask for a backup; ordinary searches show the first-choice tools as before.

## 2026-11-17: finding the right tool: a ranked search with a verdict, and tools grouped into jobs

- **[`discover_tools`](https://parel.si/docs/mcp.md#discover_tools) understands a few words** ("work email of a person", "competitor ads", "who is hiring", some Hinglish), not only an exact phrase. At most eight tools come back, best first, with a `verdict`: `match`, `nearest` (only something close) or `none` (with a hint to ask differently). On a match the first three include their input schema.
- **With no keyword it lists the jobs** Parel covers (268 tools in 35 jobs), each with its count and three examples, instead of every tool; pass `job` for one job's tools. Rows gain `job` and, on look-alike tools, `pick_when`.
- **REST**: [`POST /v1/discover`](https://parel.si/docs/api/discover.md) and [`GET /v1/tools`](https://parel.si/docs/api/tools.md) take `query`, `job` and `limit`; without them they still list every tool, now with the jobs.
- **[skill.md](https://parel.si/skill.md) is shorter** and lists the jobs; [llms.txt](https://parel.si/llms.txt) lists the tools by job.

## 2026-11-17: one account and one welcome credit per email inbox

- **One account per email inbox.** Addresses that reach the same inbox now count as one: capital letters and anything after a `+` never make a new address, and neither do dots in a Gmail address. Signing up with another spelling of an address that already has an account is refused with "An account already exists for this email address. Sign in instead." The ₹100 welcome credit is given once per inbox; accounts that already had it keep it.
- **Temporary email addresses cannot sign up.** The list of throwaway providers grew from 31 to about ten thousand.
- **Up to ten new accounts per IP address a day**, up from three. See [Limits](https://parel.si/docs/limits.md).

## 2026-11-17: six more tools: Meta ads by keyword, a website crawl, Google Shopping, flights, TikTok Shop and Airbnb listings: 191 in all

- **Six tools switched on** after a live call each: [`search_meta_ads`](https://parel.si/docs/tools/search_meta_ads.md) (ads on Facebook and Instagram that match a keyword, across every advertiser; [`get_meta_ads`](https://parel.si/docs/tools/get_meta_ads.md) still reads one brand's ads), [`crawl_site`](https://parel.si/docs/tools/crawl_site.md) (up to twenty pages of a website in one call, priced per page returned), [`search_google_shopping`](https://parel.si/docs/tools/search_google_shopping.md) (prices and sellers for a product, with `gl: "in"` for India), [`search_flights`](https://parel.si/docs/tools/search_flights.md) (Google Flights options and the typical fare for a route), [`search_tiktok_shop`](https://parel.si/docs/tools/search_tiktok_shop.md) (TikTok Shop products, the US shop in practice) and [`get_airbnb_listing`](https://parel.si/docs/tools/get_airbnb_listing.md) (one listing's details, no price). Each has a page here and a prompt on the Examples page; a call that fails or finds nothing is ₹0, as everywhere.

## 2026-11-05: seventy-six more tools: AI answers measured, TikTok, Google images, patents, hotels and quotes, deeper Instagram, YouTube, X, Reddit, Facebook, Threads, Snapchat, Pinterest, LinkedIn, weather, feeds, apps and crypto: 185 in all

- **Seventy-six tools switched on** after a live call each, in twelve families: **AI answers measured**: [`check_claude`](https://parel.si/docs/tools/check_claude.md), [`get_ai_mentions`](https://parel.si/docs/tools/get_ai_mentions.md), [`get_ai_mentioned_brands`](https://parel.si/docs/tools/get_ai_mentioned_brands.md), [`search_ai_mentions`](https://parel.si/docs/tools/search_ai_mentions.md), [`get_ai_search_volume`](https://parel.si/docs/tools/get_ai_search_volume.md), [`get_sentiment`](https://parel.si/docs/tools/get_sentiment.md); **Google beyond the ten links**: [`get_google_ads`](https://parel.si/docs/tools/get_google_ads.md), [`search_google_images`](https://parel.si/docs/tools/search_google_images.md), [`search_patents`](https://parel.si/docs/tools/search_patents.md), [`search_hotels`](https://parel.si/docs/tools/search_hotels.md), [`get_place`](https://parel.si/docs/tools/get_place.md), [`get_finance_quote`](https://parel.si/docs/tools/get_finance_quote.md); **SEO**: [`get_search_intent`](https://parel.si/docs/tools/get_search_intent.md), [`get_keyword_history`](https://parel.si/docs/tools/get_keyword_history.md), [`get_serp_competitors`](https://parel.si/docs/tools/get_serp_competitors.md), [`compare_domain_keywords`](https://parel.si/docs/tools/compare_domain_keywords.md), [`audit_page`](https://parel.si/docs/tools/audit_page.md), [`find_sites_using`](https://parel.si/docs/tools/find_sites_using.md); **Amazon**: [`get_amazon_ranked_keywords`](https://parel.si/docs/tools/get_amazon_ranked_keywords.md), [`get_amazon_competitors`](https://parel.si/docs/tools/get_amazon_competitors.md), [`get_amazon_sellers`](https://parel.si/docs/tools/get_amazon_sellers.md); **Instagram**: [`search_instagram`](https://parel.si/docs/tools/search_instagram.md), [`search_instagram_reels`](https://parel.si/docs/tools/search_instagram_reels.md), [`get_instagram_similar_profiles`](https://parel.si/docs/tools/get_instagram_similar_profiles.md), [`get_instagram_followers`](https://parel.si/docs/tools/get_instagram_followers.md), [`get_instagram_transcript`](https://parel.si/docs/tools/get_instagram_transcript.md); **YouTube**: [`get_youtube_playlists`](https://parel.si/docs/tools/get_youtube_playlists.md), [`get_youtube_community_posts`](https://parel.si/docs/tools/get_youtube_community_posts.md), [`search_youtube_hashtag`](https://parel.si/docs/tools/search_youtube_hashtag.md), [`get_youtube_trending_shorts`](https://parel.si/docs/tools/get_youtube_trending_shorts.md); **TikTok**: [`search_tiktok`](https://parel.si/docs/tools/search_tiktok.md), [`get_tiktok_profile`](https://parel.si/docs/tools/get_tiktok_profile.md), [`get_tiktok_videos`](https://parel.si/docs/tools/get_tiktok_videos.md), [`get_tiktok_video`](https://parel.si/docs/tools/get_tiktok_video.md), [`get_tiktok_comments`](https://parel.si/docs/tools/get_tiktok_comments.md), [`get_tiktok_transcript`](https://parel.si/docs/tools/get_tiktok_transcript.md), [`search_tiktok_hashtag`](https://parel.si/docs/tools/search_tiktok_hashtag.md), [`get_tiktok_trending`](https://parel.si/docs/tools/get_tiktok_trending.md), [`get_tiktok_trending_hashtags`](https://parel.si/docs/tools/get_tiktok_trending_hashtags.md); **X**: [`get_x_followers`](https://parel.si/docs/tools/get_x_followers.md), [`get_x_following`](https://parel.si/docs/tools/get_x_following.md), [`get_x_trending`](https://parel.si/docs/tools/get_x_trending.md), [`get_x_thread`](https://parel.si/docs/tools/get_x_thread.md), [`search_x_users`](https://parel.si/docs/tools/search_x_users.md); **Reddit**: [`get_reddit_post`](https://parel.si/docs/tools/get_reddit_post.md), [`get_reddit_user_posts`](https://parel.si/docs/tools/get_reddit_user_posts.md), [`get_reddit_user_comments`](https://parel.si/docs/tools/get_reddit_user_comments.md), [`get_reddit_user`](https://parel.si/docs/tools/get_reddit_user.md), [`get_reddit_trending`](https://parel.si/docs/tools/get_reddit_trending.md), [`get_reddit_popular`](https://parel.si/docs/tools/get_reddit_popular.md); **Facebook, Threads, Snapchat, Pinterest**: [`get_facebook_post`](https://parel.si/docs/tools/get_facebook_post.md), [`get_facebook_page`](https://parel.si/docs/tools/get_facebook_page.md), [`search_facebook_events`](https://parel.si/docs/tools/search_facebook_events.md), [`get_facebook_events`](https://parel.si/docs/tools/get_facebook_events.md), [`search_threads`](https://parel.si/docs/tools/search_threads.md), [`get_threads_posts`](https://parel.si/docs/tools/get_threads_posts.md), [`get_threads_profile`](https://parel.si/docs/tools/get_threads_profile.md), [`get_threads_post`](https://parel.si/docs/tools/get_threads_post.md), [`get_snapchat_profile`](https://parel.si/docs/tools/get_snapchat_profile.md), [`search_pinterest`](https://parel.si/docs/tools/search_pinterest.md); **LinkedIn**: [`get_linkedin_post`](https://parel.si/docs/tools/get_linkedin_post.md), [`get_linkedin_reactions`](https://parel.si/docs/tools/get_linkedin_reactions.md), [`get_linkedin_comments`](https://parel.si/docs/tools/get_linkedin_comments.md), [`search_linkedin_companies`](https://parel.si/docs/tools/search_linkedin_companies.md); **Everyday and companies**: [`get_weather`](https://parel.si/docs/tools/get_weather.md), [`read_feed`](https://parel.si/docs/tools/read_feed.md), [`get_medium_posts`](https://parel.si/docs/tools/get_medium_posts.md), [`get_substack_posts`](https://parel.si/docs/tools/get_substack_posts.md), [`search_indeed_jobs`](https://parel.si/docs/tools/search_indeed_jobs.md), [`get_app_store_app`](https://parel.si/docs/tools/get_app_store_app.md), [`get_google_play_app`](https://parel.si/docs/tools/get_google_play_app.md), [`get_github_user_repos`](https://parel.si/docs/tools/get_github_user_repos.md), [`get_crypto_prices`](https://parel.si/docs/tools/get_crypto_prices.md), [`get_email_pattern`](https://parel.si/docs/tools/get_email_pattern.md), [`find_company_domain`](https://parel.si/docs/tools/find_company_domain.md), [`find_company_linkedin`](https://parel.si/docs/tools/find_company_linkedin.md). Each has a page here and a prompt on the Examples page. The TikTok pages say plainly that TikTok is blocked in India and read the markets you sell into. The AI-mentions tools cost about ten times a search and say so on their pages; `find_company_domain` is free.

## 2026-11-05: eighteen more tools: creators, news by meaning, long lists, Telegram, Facebook groups, GitHub, Bing, a free phone check: 109 in all

- **Eighteen tools switched on** after a live call each: [`search_creators`](https://parel.si/docs/tools/search_creators.md), [`search_news_semantic`](https://parel.si/docs/tools/search_news_semantic.md), [`search_deep_agent`](https://parel.si/docs/tools/search_deep_agent.md), [`list_companies_by_description`](https://parel.si/docs/tools/list_companies_by_description.md) and [`list_people_by_description`](https://parel.si/docs/tools/list_people_by_description.md) (up to 100 for one flat price), [`get_telegram_channel_posts`](https://parel.si/docs/tools/get_telegram_channel_posts.md), [`get_telegram_channel`](https://parel.si/docs/tools/get_telegram_channel.md) and [`search_telegram_channel`](https://parel.si/docs/tools/search_telegram_channel.md), [`get_facebook_group_posts`](https://parel.si/docs/tools/get_facebook_group_posts.md) and [`get_facebook_group`](https://parel.si/docs/tools/get_facebook_group.md), [`get_youtube_shorts`](https://parel.si/docs/tools/get_youtube_shorts.md), [`get_subreddit`](https://parel.si/docs/tools/get_subreddit.md), [`get_github_repo`](https://parel.si/docs/tools/get_github_repo.md), [`get_github_user`](https://parel.si/docs/tools/get_github_user.md) and [`get_github_trending`](https://parel.si/docs/tools/get_github_trending.md), [`search_bing`](https://parel.si/docs/tools/search_bing.md), [`get_wikipedia_article`](https://parel.si/docs/tools/get_wikipedia_article.md) and [`validate_phone`](https://parel.si/docs/tools/validate_phone.md), which is free. Each has a page here and a prompt on the Examples page.

## 2026-11-05: LinkedIn jobs, Glassdoor reviews and Crunchbase profiles: 91 in all

- **Three tools switched on**: [`search_linkedin_jobs`](https://parel.si/docs/tools/search_linkedin_jobs.md) (LinkedIn's public job search by keyword and location), [`get_glassdoor_reviews`](https://parel.si/docs/tools/get_glassdoor_reviews.md) (a company's employee reviews, by its Glassdoor slug) and [`get_crunchbase_company`](https://parel.si/docs/tools/get_crunchbase_company.md) (a startup's public Crunchbase profile with funding rounds and investors). Their source succeeds less often than the rest of the catalogue (93%, 93% and 64% on the upstream's numbers); each page says so, and a failed run is ₹0.

## 2026-11-05: wave 2, nine everyday and competitor tools: 88 in all

- **Nine tools switched on** after a live call each: [`get_facebook_posts`](https://parel.si/docs/tools/get_facebook_posts.md) and [`get_facebook_comments`](https://parel.si/docs/tools/get_facebook_comments.md) (a Page's posts, a post's comments), [`search_linkedin_ads`](https://parel.si/docs/tools/search_linkedin_ads.md) (a company's ads in the LinkedIn Ad Library), [`get_funding_news`](https://parel.si/docs/tools/get_funding_news.md) (who just raised, by country, priced per round), [`search_scholar`](https://parel.si/docs/tools/search_scholar.md) (research papers), [`search_wikipedia`](https://parel.si/docs/tools/search_wikipedia.md), [`search_hacker_news`](https://parel.si/docs/tools/search_hacker_news.md) and [`get_product_hunt_launches`](https://parel.si/docs/tools/get_product_hunt_launches.md), and [`verify_phone`](https://parel.si/docs/tools/verify_phone.md) (is a number valid, and what kind). Each has a page here and a prompt on the Examples page.

## 2026-11-05: a tool that decides: 79 in all

- **[`judge`](https://parel.si/docs/tools/judge.md) switched on** after live calls: a yes/no probability, a label or a score over evidence you supply, for sorting and filtering many records. It decides; it does not write. The evidence goes to an outside AI model, and the [Privacy Notice](https://parel.si/legal/privacy) now says so.


## 2026-11-05: nine tools for what people say and what stores sell: 78 in all

- **Nine tools switched on** after live calls: YouTube transcripts and a channel's latest videos; Instagram comments and hashtag posts; LinkedIn post search, comments and reactions; replies to an X post; and a public Shopify store's products. A page with nothing on it is ₹0.


## 2026-11-05: a known person's phone number: 69 in all

- **[`find_phone`](https://parel.si/docs/tools/find_phone.md) switched on** after live calls: the verified mobile number of one known person, from a LinkedIn URL, work email, or name and employer. It charges only when a number is found, and most people have none, which is ₹0. One person per call, no reverse lookup, no bulk lists. The number is not stored by Parel.
- **Acceptable Use Policy** version 0.3: section 1 now covers phone numbers.


## 2026-11-05: company and person details: 68 in all

- **Two tools switched on** after live calls: [`enrich_company`](https://parel.si/docs/tools/enrich_company.md) gives a company's industry, size, location and links from its domain, and [`enrich_person`](https://parel.si/docs/tools/enrich_person.md) gives a known person's title, employer, seniority and location. Both charge only when something is found. `enrich_person` takes a work address, never a personal one, and returns no email or phone number.


## 2026-11-05: four more search tools: 66 in all

- **Four tools switched on** after live calls: [`search_podcasts`](https://parel.si/docs/tools/search_podcasts.md) finds podcast episodes by what was said in them, [`search_deep`](https://parel.si/docs/tools/search_deep.md) is a more thorough web search that returns the passages answering a question, and [`search_companies_by_description`](https://parel.si/docs/tools/search_companies_by_description.md) and [`search_people_by_description`](https://parel.si/docs/tools/search_people_by_description.md) find companies and people from a plain-language description.


## 2026-11-05: screenshot pictures on Parel's own address

- **[`screenshot_page`](https://parel.si/docs/tools/screenshot_page.md) now stores the picture**: `data.image` is a link under `api.parel.si` that works for 30 days, and `data.image_expires_at` says until when. Before, the link pointed at a third-party host and was not kept for long. A picture that could not be stored is a free, retryable error.

## 2026-11-04: three web tools for lists that span pages: 62 in all

- **Three tools switched on** after a live call each: [`map_site`](https://parel.si/docs/tools/map_site.md) lists every URL on a website, or under one path, in one call, so an agent finds the pages of a list before reading any of them; [`extract_structured`](https://parel.si/docs/tools/extract_structured.md) turns one page into the records you describe (a table of names and titles, prices, any fields) from a sentence or a schema, shown at the most a page can cost and settled at its real cost; [`screenshot_page`](https://parel.si/docs/tools/screenshot_page.md) returns a real-browser screenshot of a page as a link to a PNG (since 2026-11-05 on Parel's own address).
- **`read_page` now points the way**: its description tells an agent that a list spanning pages is `map_site`, then `read_pages` (ten URLs a call) or `extract_structured`, not one `read_page` call per page.

## 2026-10-28: twenty-two more tools, three of them free: 59 in all

- **Twenty-two tools switched on** after a live call each: X search, profile and post; people search and people count (free); job postings and one LinkedIn job; a company page's posts, its technology stack and lookalike companies (free); Google's local pack, Google AI Mode and Gemini answers; a domain's overview, its backlink and referring-domain lists and authority score; the comments under a Reddit post; one Instagram post and an account's reels; a page as typed fields; App Store search. Each has a page here and a prompt on the Examples page.
- **Three tools are free** (`search_people`, `count_people`, `find_similar_companies`): every call answers `class: "ok_free"` with `cost_paise: 0`.

## 2026-10-28: seventeen more tools: LinkedIn, companies, Amazon, Meta ads, more SEO and social

- **Seventeen tools switched on** after a live call each, bringing the catalogue to 37: `get_linkedin_profile`, `search_linkedin_people`, `get_linkedin_posts`, `get_linkedin_company`, `search_companies`, `answer_question`, `read_pages`, `get_meta_ads`, `get_keyword_ideas`, `get_ranked_keywords`, `get_maps_reviews`, `search_amazon`, `get_amazon_product`, `search_reddit`, `search_youtube`, `get_youtube_channel` and `search_instagram_users`. Each has a page here with inputs, today's price and a real recorded answer; the Examples page has a prompt for each.
- **Prices that depend on what you ask for** say so: `search_companies` (per row), `read_pages` (per page per content type), `get_keyword_ideas` and `get_ranked_keywords` (per keyword returned) show the price of the largest call and settle lower for a smaller one.

## 2026-10-28: twelve more tools, five better sources

- **Twelve tools switched on** after a live check against the source the same day: `search_maps`, `search_semantic`, `rank_check`, `keyword_data`, `keyword_volume` (new: Google Ads monthly volumes and cost per click for up to a thousand keywords in one call), `backlinks`, `domain_search`, `get_instagram_profile`, `get_youtube`, `get_youtube_comments`, `get_x_posts` and `check_perplexity`. Twenty tools in all; each has a page here with its inputs, today's price and a real recorded answer.
- **Five tools moved to a better source**, with the fields you send unchanged: `verify_email` (a live mailbox check at a fifth of the price, with `validity` and a reason instead of a score), `search_maps`, `get_instagram_profile` and `get_x_posts` (cheaper, smaller, faster answers; a missing account is ₹0) and `check_perplexity` (the planned source could not reach Perplexity; the new one answers in five seconds with sources). `verify_email`'s answer changed shape: `data.validity` instead of `data.data.status`.
- **Prices made honest**: `search_semantic` is flat per call for up to ten results (it was shown per result). The site's prices refreshed to today's rate.

## 2026-10-28: one account, no teams to set up

- Signing up no longer asks you to name a team. Your account is created the moment your email is verified, with the welcome credit, and you go straight to connecting Claude.
- The Members page and invitations are off during the closed beta: every account is a team of one. The API answers `403 teams_hidden` to invitations for now; the `/account/teams/...` paths and `team_id` fields are unchanged.

## 2026-10-28

- **Documentation rebuilt** as this section: a page per endpoint, per MCP tool and per data tool, every example captured from staging, a Markdown twin of every page, `llms.txt` rewritten for agents and `llms-full.txt` added. Prices on the site refreshed to today's rate.
- **Legal pages, version 0.2**: Terms, Privacy Notice, Refund Policy, Acceptable Use, Security and Data Processing Terms rewritten in plain English against the product's actual rules and Indian law as of 2026, each with a short version at the top.
- **Empty balance during the beta**: the `insufficient_balance` error and `check_balance` now say how to get credit (email hello@parel.si) instead of pointing at a top-up page that is off.

## 2026-10-19

- **Sign in from Claude**: "Add custom connector" on claude.ai works with URL + Sign in, no key to paste. A signed-in person gets a hidden "Claude connector" key on their team, listed under API keys; revoking it disconnects Claude.
- **Tool quality**: hand-written descriptions for every tool (what it returns, what it will not do, which sibling to use, what drives the cost, India notes), notes on quirks and a valid example input; `inspect_tool` and `GET /v1/tools/:id` show them.
- **Confirmation step**: any call estimated above ₹200 answers `confirmation_required` (HTTP 428, ₹0) with the exact quote until the same input is resent with `confirm: true`.
- **Published error table**: every error code with what you pay and whether to retry; `X-Request-Id` on every response and `request_id` in every error body.
- **Discover**: headline tools first, summaries only in long lists, disabled tools never listed; search also matches summaries.
- **The website** at parel.si with docs, status and legal pages, and the dashboard brought to launch quality: onboarding, keys with caps in rupees, a usage log with plain-English reasons under failed calls, members and invitations.
- **The eight beta tools** switched on: `search_web`, `search_news`, `read_page`, `find_email`, `verify_email`, `get_social_posts`, `get_reddit`, `check_ai_visibility`.

## 2026-10-17

- **Payments with tax**: a top-up that includes GST on top of the pack price is credited rather than rejected. (Purchases are still off during the beta.)
- **Request bodies** limited to 1 MB (`body_too_large`, HTTP 413).
- **Welcome credit protected**: no credit for throwaway email domains, with daily limits on new teams.

## 2026-10-16

- **Prices from the live rate**: every rupee figure an agent sees is computed by the same formula and exchange rate that settle the call.
- **Free quotes**: `POST /v1/quote` and `quote_tool` return the charge and hold for an exact input, with nothing recorded or held.
- **One-of inputs**: tools can declare alternative field sets; `find_email` refuses a lookup with no name before any upstream call, for ₹0.
- **`inspect_tool`** added; `discover_tools` includes the schema for short lists.
- **Large answers over MCP** cut to whole list items (valid JSON, with a `_truncated` note) instead of mid-object; billing unchanged, REST returns everything.
- **The dashboard**: sign up, create a team with ₹100 of credit, create and revoke keys, read the usage log where every failed call reads "failed, charged ₹0", report a bad result, set spend caps, invite colleagues.
- **`skill.md` and `llms.txt`** written for real.
