This timeline tracks the moments when tech and AI companies restricted, revoked, paywalled or blocked each other's APIs, data, crawlers, apps or agents, plus the lawsuits, reversals and licensing deals that settled those fights. The "access layer" has moved three times: first APIs (who gets the firehose), then crawlers (who gets to train), now agents and MCP (who gets to act for a user).The trigger was a September 30, 2026 Wes Bos post quote-posting Figma CEO Dylan Field's reply to MCP co-creator David Soria Parra: "Figma doesn't want you to use MCP. Amazon doesn't want you to shop with agents. Reddit won't let Claude access threads. X won't let ChatGPT read tweets. It's starting. This happened with APIs 8 years ago." Every entry below links to the companies' own statements or to coverage from the time. Entries marked (secondary source) still need checking against a primary document before publication.Era 1: The Open-API Boom and the First Walls (2007–2013)#1 | May 24, 2007: Facebook Platform launches. Facebook opens its social graph to outside developers with 65 partners and 85 apps. That openness becomes the template, and the target, for every later lockdown.#2 | December 2008: Facebook sues Power Ventures. Facebook sues the aggregator Power.com for logging into users' accounts on their behalf after a cease-and-desist. The Ninth Circuit later holds that access after the letter violated the CFAA.#3 | August 31, 2010: Twitter shuts off Basic Auth. The "OAuthpocalypse" forces every third-party app onto OAuth or off the platform. It is the first big signal that Twitter will decide how apps reach its API.#4 | November 2010: Google cuts Facebook's Gmail Contacts API access. Google bars sites that don't allow similar export from automatically importing Gmail contacts. It is the first public data-reciprocity fight between two giants.#5 | March 11, 2011: Twitter tells developers not to build client apps. Platform lead Ryan Sarver says developers should not build apps that mimic the core Twitter experience, warning off the ecosystem that made Twitter popular.#6 | May 26, 2011: Google deprecates the free Translate API. Google cites "extensive abuse" and sets a shutdown date, then relaunches Translate as a paid API after a backlash. Free-to-paid becomes a pattern.#7 | June 29, 2012: Twitter stops syncing tweets to LinkedIn. Tweets stop flowing into LinkedIn feeds as Twitter pulls its content back onto its own surfaces.#8 | July 2012 to August 2013: Craigslist wins its IP-block ruling against 3Taps. Craigslist blocks 3Taps' IP addresses and sues. On August 16, 2013, a judge rules that scraping after an IP block and cease-and-desist can violate the CFAA.#9 | July 26, 2012: Twitter cuts Instagram's Find Friends. After Facebook buys Instagram, Twitter removes Instagram's access to its follow graph.#10 | August 16, 2012: Twitter's API v1.1 caps clients at 100,000 users. New token caps and display requirements put a ceiling on apps like Tweetbot and Echofon.#11 | Late summer 2012: Twitter cuts Tumblr's Find Friends. Another rival loses access to Twitter's social graph.#12 | December 5, 2012: Instagram pulls its Twitter Cards. Instagram photos stop displaying inside tweets, so users have to click through to Instagram.#13 | January 24, 2013: Facebook blocks Vine's friend-finding on launch day. Facebook cuts Twitter's new video app off its friend data within hours of launch, citing its rule against replicating core functionality.#14 | January 2013: Facebook cuts Voxer from Find Friends. The messaging app loses friend-finding the same month, along with Yandex's Wonder.#15 | March 13, 2013: Google announces the Reader shutdown. Reader closes on July 1, 2013, ending the backend that many RSS clients relied on.#16 | August 2013: Google blocks Microsoft's Windows Phone YouTube app. Google disables Microsoft's rebuilt app for violating YouTube's terms; Microsoft calls the reasons "manufactured."Era 2: Platform Consolidation and the Graph Lockdown (2014–2017)#17 | April 30, 2014: Facebook's Graph API 2.0 ends friends-data access. New apps can no longer pull friends' data; existing apps get a one-year grace period. Cambridge Analytica's data collection ran in that window.#18 | June 13, 2014: Netflix announces the end of its public API. Public developers lose access on November 14, 2014, and every request returns a 404. Only select partners keep access.#19 | February 12, 2015: LinkedIn restricts its APIs to partners. From May 12, 2015, only four open uses remain; everything else requires a partner program.#20 | March 13, 2015: Twitter cuts Meerkat off its social graph. As Twitter closes its Periscope purchase, it removes the rival live-streaming app's ability to port Twitter follows.#21 | April 30, 2015: Graph API v1.0 is switched off. The friends-permissions grace period ends for all apps.#22 | June to December 2015: Twitter revokes, then restores, Politwoops. Twitter cuts the Sunlight Foundation's deleted-tweet archive off its API, then restores access on December 31, 2015, an early access reversal.#23 | November 17, 2015: Instagram's Platform Policy lockdown. From June 1, 2016, Instagram's API is limited to approved use cases, ending third-party feed readers like Retro and Flow.#24 | May 2016: Twitter cuts U.S. spy agencies off Dataminr. Twitter stops Dataminr, which holds firehose access, from selling real-time alerts to U.S. intelligence agencies.#25 | September to October 2016: Facebook, Instagram and Twitter cut off Geofeedia. After an ACLU report shows police used the tool to monitor protesters, all three platforms end its data access.#26 | August 14, 2017: Court orders LinkedIn to unblock hiQ Labs. LinkedIn sent hiQ a cease-and-desist in May 2017; a judge orders LinkedIn to remove its technical blocks within 24 hours.#27 | December 5, 2017: Google pulls YouTube from Amazon's Echo Show and Fire TV. Google cites a "lack of reciprocity" because Amazon won't sell Chromecast or Google Home. The two reach a truce in 2019.Era 3: The Cambridge Analytica Lockdown, the "8 Years Ago" Moment (2018–2019)#28 | March 16, 2018: Facebook suspends Cambridge Analytica. Facebook suspends the firm as reports reveal data on up to 87 million users was harvested through a quiz app.#29 | April 4, 2018: Facebook restricts its Events, Groups and Pages APIs. Apps lose group member lists; all future Pages API access needs Facebook's approval.#30 | April 2018: Instagram slashes its API rate limits. Without notice, Instagram cuts calls from 5,000 to 200 per user per hour and speeds up its legacy API's retirement.#31 | July 24, 2018: Twitter removes 143,000 apps and gates its API. Every new API request must now pass a developer-account review, capped at 10 apps per account.#32 | August 16, 2018: Twitter retires User Streams. Real-time timelines and push notifications break in Tweetbot and Twitterrific as Twitter pushes developers to its paid Account Activity API.#33 | October 8, 2018: Project Strobe and the Google+ shutdown. After a bug exposed profile data, Google kills consumer Google+ and restricts Gmail API access behind security assessments.#34 | January 30, 2019: Apple revokes Facebook's enterprise certificate. Apple pulls the certificate over Facebook's "Research" VPN app, breaking its internal apps; Google's is pulled next over Screenwise Meter.#35 | July 24, 2019: FTC's $5 billion Facebook order. The record privacy penalty puts the post-2018 data lockdown under 20 years of federal oversight.Era 4: Gatekeepers, App Stores and Scraping Law (2020–2022)#36 | August 13–28, 2020: Apple pulls Fortnite and terminates Epic's developer account. Epic adds direct payments to dodge Apple's cut; Apple removes the game and cuts Epic off, starting Epic v. Apple.#37 | January 9–10, 2021: AWS, Apple and Google cut off Parler. App-store removals and an AWS hosting termination take the service offline, showing infrastructure access can vanish overnight.#38 | April 2021 to February 2022: App Tracking Transparency costs Meta about $10 billion. Apple's iOS 14.5 prompt limits apps' access to the ad identifier; Meta's CFO calls the 2022 hit "on the order of $10 billion."#39 | August 3, 2021: Facebook disables NYU Ad Observatory accounts. Facebook cuts off the political-ad researchers' accounts and API access, citing scraping and its FTC order.#40 | April 18, 2022: Ninth Circuit reaffirms hiQ v. LinkedIn. The court holds that scraping a public website likely isn't "without authorization" under the CFAA, shaping the crawler fights that follow.#41 | November 30, 2022: ChatGPT launches. Public web text becomes the most valuable training input in tech, and every platform reprices its data.Era 5: The Generative-AI Data Wars (2023–2024)#42 | January 19, 2023: Twitter bans third-party clients. A week after Tweetbot and Twitterrific stop working, Twitter adds a developer-terms clause banning apps that substitute for its own.#43 | February to March 2023: Twitter's $42,000-a-month enterprise API. Twitter ends its free API; Wired reports enterprise tiers starting at $42,000 a month for 50 million tweets.#44 | April 7, 2023: Twitter throttles Substack links. Days after Substack announces Notes, tweets with Substack links can't be liked or retweeted; the limit lifts on April 8.#45 | April 18, 2023: Reddit says AI companies must pay for its API. CEO Steve Huffman tells The New York Times that Reddit won't give its data "to some of the largest companies in the world for free."#46 | April 19, 2023: Musk's "lawsuit time" threat at Microsoft. After Microsoft drops Twitter from its ad platform, Musk accuses it of training illegally on Twitter data.#47 | April 20, 2023: Stack Overflow plans to charge AI developers. CEO Prashanth Chandrasekar says LLM builders should pay for its 50 million questions and answers.#48 | June 2023: Reddit's API pricing kills Apollo. At $12,000 per 50 million requests, developer Christian Selig says Apollo would owe about $20 million a year; it shuts June 30.#49 | June 14, 2023: Musk says OpenAI has no authorized Twitter feed. After users show ChatGPT browsing reading tweets, Musk publicly rules it out. This is the origin of the "X won't let ChatGPT read tweets" storyline.#50 | July 1, 2023: Twitter's login wall and daily reading limits. Musk caps unverified accounts at 600 posts a day to address "extreme levels of data scraping & system manipulation."#51 | July 2023 to May 2024: X sues Bright Data, and loses. A judge dismisses the scraping suit in May 2024, warning against "information monopolies."#52 | August 3, 2023: The New York Times bans AI training in its terms. The new terms bar using Times content to train any machine-learning system without written permission.#53 | August 7, 2023: OpenAI publishes its GPTBot opt-out. Within weeks, the NYT, CNN, Reuters and others add GPTBot to their robots.txt disallow lists.#54 | September 8, 2023: X bans crawling without written consent. X's new terms drop the old allowance for robots.txt-compliant crawling.#55 | September 28, 2023: Google launches the Google-Extended opt-out. Publishers can block Bard and Vertex AI training without leaving Search; there is no separate opt-out for AI search features.#56 | End of 2023: 48% of top news sites block OpenAI's crawlers. The Reuters Institute finds the share ranges from 79% in the U.S. to 20% in Mexico and Poland.#57 | December 27, 2023: The New York Times sues OpenAI and Microsoft. The suit says the companies used millions of Times articles without permission and caused "billions of dollars" in damages.#58 | January 23, 2024: Meta loses its Bright Data scraping case. A judge rules logged-out scraping of public Facebook and Instagram data didn't breach Meta's terms; Meta drops the case a month later.#59 | February 22, 2024: Reddit and Google sign a Data API deal. Google gets real-time access to Reddit's Data API for training and display, reportedly about $60 million a year.#60 | May 6, 2024: Stack Overflow and OpenAI partner. The data fight settles into a license: OpenAI uses Stack Overflow's OverflowAPI.#61 | May 16, 2024: Reddit and OpenAI sign a data deal. OpenAI gets Data API access and becomes a Reddit advertiser; terms aren't disclosed.#62 | June 2024: Applebot-Extended opt-out. Apple gives publishers a training opt-out; by late August, Facebook, Instagram, the NYT, the FT, Condé Nast and others use it.#63 | June 19, 2024: Wired calls Perplexity a "bullshit machine". Wired finds Perplexity scraping pages its crawler was blocked from; Forbes had accused it of plagiarism earlier that month.#64 | June 25, 2024: Reddit's robots.txt shuts out unlicensed crawlers. Within weeks, Reddit results vanish from Bing and DuckDuckGo while Google, a licensee, keeps them.#65 | July 3, 2024: Cloudflare ships one-click AI-bot blocking. Cloudflare says AI bots reached 39% of its top million sites in June, but under 3% blocked them.#66 | August 14, 2024: Meta shuts down CrowdTangle. Researchers lose their main tool for monitoring public Facebook and Instagram content ahead of the U.S. election.#67 | September 18, 2024: LinkedIn trains AI on user data before updating its terms. Users outside the EU, EEA and Switzerland are opted in by default.#68 | October 21, 2024: Dow Jones and the New York Post sue Perplexity. The suit lands days after the NYT sends Perplexity a cease-and-desist over scraping.#69 | November 25, 2024: Anthropic launches the Model Context Protocol. MCP becomes the open standard for connecting agents to tools and data, and the next layer where access fights happen.#70 | December 4, 2024: Instagram's Basic Display API shuts down. Personal accounts lose all third-party API access; only business and creator accounts remain.Era 6: Agents, MCP and the Access-Layer War (2025 to October 1, 2026)#71 | January 29, 2025: OpenAI accuses DeepSeek of distilling its models. OpenAI and Microsoft review and block accounts they link to DeepSeek; model output becomes guarded territory.#72 | February 24, 2025: Chegg sues Google over AI Overviews. The first single-company antitrust suit over AI summaries alleges Google forces publishers to feed its answers.#73 | March 19, 2025: Cloudflare launches AI Labyrinth. Crawlers that ignore no-crawl rules get trapped in mazes of AI-generated decoy pages.#74 | May 29, 2025: Salesforce locks down Slack API data. New terms bar bulk export, persistent indexes and LLM training on Slack API data.#75 | June 3, 2025: Anthropic cuts Windsurf's Claude capacity. Amid reports OpenAI would buy Windsurf, co-founder Jared Kaplan says it "would be odd for us to be selling Claude to OpenAI."#76 | June 4, 2025: Reddit sues Anthropic. Reddit alleges Anthropic's bots hit the site more than 100,000 times after Anthropic said they'd stopped.#77 | June 4, 2025: X bans training foundation models on its content. The developer agreement now forbids using the X API or X content to "fine-tune or train a foundation or frontier model."#78 | July 1, 2025: Cloudflare blocks AI crawlers by default. New domains are opted out of AI crawling unless they allow it, and publishers can charge through Pay Per Crawl.#79 | About August 1, 2025: Anthropic revokes OpenAI's Claude API access. Anthropic says OpenAI staff used Claude Code ahead of GPT-5 in violation of its terms; benchmarking access stays.#80 | August 4, 2025: Cloudflare accuses Perplexity of stealth crawling. Cloudflare says Perplexity rotated user agents and networks to evade blocks, and delists it as a verified bot.#81 | September 5, 2025: Anthropic's $1.5 billion authors settlement. About $3,000 per book for some 500,000 pirated books; the authors' lawyer calls it the largest copyright recovery ever.#82 | September 12, 2025: Penske Media sues Google over AI Overviews. The Rolling Stone publisher says Google only indexes sites that also let it use their articles in AI summaries.#83 | October 21, 2025: ChatGPT Atlas presents itself as plain Chrome. Observers note OpenAI's agentic browser is hard to block or track. OpenAI later folds Atlas into ChatGPT.#84 | October 22, 2025: Reddit sues Perplexity, SerpApi and two proxy firms. Reddit alleges "industrial-scale" scraping of its content through Google results.#85 | November 4, 2025: Amazon sues Perplexity over Comet. Amazon says the shopping agent disguised itself as Chrome inside customer accounts; Perplexity calls the suit bullying.#86 | Late 2025: Amazon blocks ChatGPT's shopping crawlers. OpenAI's agents join Anthropic, Perplexity and Google bots on Amazon's robots.txt block list.#87 | December 19, 2025: Google sues SerpApi. Google alleges hundreds of millions of daily fake searches that dodge its SearchGuard protections.#88 | January 9, 2026: Anthropic blocks Claude subscription logins in third-party harnesses. Server-side checks stop OpenCode and others from using Claude Pro and Max tokens outside Claude Code.#89 | January 15, 2026: X's new terms add scraping damages. The terms reportedly add $15,000 in damages per million posts accessed in a day. (secondary source)#90 | February 17, 2026: Figma and Anthropic launch Code to Canvas. Work from Claude Code can be sent into editable Figma designs, the partnership's high-water mark.#91 | February 19–20, 2026: Anthropic writes the harness ban into its terms. OpenCode strips Claude subscription support, citing Anthropic legal requests.#92 | March 9, 2026: Court bars Comet from Amazon accounts. Judge Maxine Chesney finds Comet entered accounts "with the Amazon user's permission, but without authorization by Amazon."#93 | March 30, 2026: Reddit v. Anthropic goes back to state court. A judge rules Reddit's contract and tort claims aren't preempted by copyright law.#94 | April 4, 2026: Claude subscriptions stop covering OpenClaw and other harnesses. Third-party agent use moves to metered billing.#95 | April 17–23, 2026: Claude Design launches and the Figma feud goes public. Days after an Anthropic executive leaves Figma's board, Dylan Field says Anthropic was "not consistently candid."#96 | 2026, ongoing: Figma's remote MCP server accepts only listed clients. Unlisted clients get 403 errors at sign-up; users call it a break with MCP's core promise.#97 | May 30, 2026: Reddit blocks anonymous JSON access. Unauthenticated requests start returning 403 errors, breaking Claude MCP servers that relied on them.#98 | June 2026: Slack opens its data to Claude. A year after its API lockdown, Salesforce invites one partner in. (secondary source)#99 | July 1, 2026: Cloudflare announces Pay Per Use. From September 15, mixed-use AI crawlers are blocked by default on new and free-plan domains, and publishers are paid when answers use their content.#100 | July 22, 2026: Reddit may not renew its Google AI deal. CNBC, citing the WSJ, reports the 2024 license may lapse.#101 | Late July 2026: Reddit's DMCA claims against Perplexity and SerpApi survive. A Manhattan judge lets the core circumvention claims proceed.#102 | July 2026: A court tosses most of Google's DMCA claims against SerpApi. The judge finds most search-result data isn't copyrighted; Google files an amended complaint on August 10.#103 | August 4, 2026: The Ninth Circuit vacates the Comet injunction. The court holds the agent "is a tool, not a person" and the user, not Perplexity, accesses Amazon; Amazon can still regulate through its terms.#104 | August 14, 2026: Amazon's Agent Terms require AI agents to identify themselves. (secondary source)#105 | August 24–25, 2026: X sends Nitter a cease-and-desist. X cites scraping and API circumvention; nitter.net goes offline.#106 | August 28, 2026: SerpApi countersues Reddit on antitrust grounds. SerpApi alleges Reddit served Google a different robots.txt file than everyone else.#107 | September 8, 2026: Meta launches its Muse shopping agent. Partners include Walmart, Sephora, Expedia and Shopify, but not Amazon.#108 | September 10, 2026: OpenAI blocks rival AI ads inside ChatGPT. Ads for competing image and audio AI tools are restricted. (secondary source)#109 | September 11, 2026: Perplexity moves to dismiss Amazon's remaining claims. (secondary source)#110 | September 14, 2026: xAI drops its Apple antitrust suit. The suit against OpenAI continues.#111 | September 15, 2026: Cloudflare's default block on mixed-use crawlers takes effect.#112 | September 18, 2026: Claude in Chrome stops working on Reddit. Every tool call fails with a "safety restrictions" message from Anthropic's own extension; the cause is unconfirmed.#113 | September 19, 2026: Reddit's suit against Anthropic survives a demurrer. Commentators write that terms of service "just got teeth."#114 | September 20, 2026: Amazon blocks Meta's Muse. Users see that "continued access by an unauthorized AI agent violates Amazon's Conditions of Use."#115 | September 29, 2026: OpenAI launches dots and Pi adds MCP. OpenAI's always-on agents run cloud browsers; Mario Zechner's Pi agent ships core MCP support after a year without it.#116 | September 30, 2026: Dylan Field asks why Anthropic resources aren't on MCP. Field answers David Soria Parra's "please fix this," and Wes Bos quote-posts: "It's starting."Fact-Check: Wes Bos's Four Claims"Figma doesn't want you to use MCP." True, with a qualification. Figma ships MCP and supports major clients, but its remote server accepts only clients in its catalog. The fight is over which clients, not whether MCP exists."Amazon doesn't want you to shop with agents." True. Amazon sued Perplexity, blocks rival AI crawlers and blocked Muse. It still runs its own Rufus and Buy for Me agents, so the accurate version is "other companies' agents.""Reddit won't let Claude access threads." Mostly true. Reddit is suing Anthropic and closed anonymous JSON access. The September 18 Claude in Chrome failure shows Anthropic's own error message, so who triggered that block is unconfirmed."X won't let ChatGPT read tweets." Plausible, not confirmed for 2026. X has banned unauthorized crawling since 2023 and Musk ruled out OpenAI's access in 2023. No new 2026 block aimed at OpenAI turned up."This happened with APIs 8 years ago." True. Facebook's April 2018 lockdown and Twitter's August 2018 streaming-API retirement are exactly eight years back.CaveatsEntries marked (secondary source) rely on trade or aggregator coverage; check them against the primary documents.The exact day Anthropic cut off OpenAI is uncertain; coverage ran around August 1, 2025.Counts of Amazon's blocked AI bots come from third-party readings of its robots.txt, not from Amazon.Dropped for lack of a source: Facebook cutting off MessageMe in 2013, an Atlassian restriction aimed at AI tools, and the June 2025 Figma MCP launch date.The Takeaway: Every Open Layer Gets a Toll BoothOpenness is a growth strategy, not a principle. Platforms open APIs, crawl access and MCP endpoints while outsiders bring them users, then close once outsiders start owning the customer relationship.These fights rarely end in court wins. They end in licenses (Reddit and Google, Stack Overflow and OpenAI), tolls (Cloudflare's Pay Per Crawl and Pay Per Use) or allowlists (Figma's MCP catalog). With the Ninth Circuit's Comet ruling weakening the anti-hacking law as a weapon, terms of service and commercial leverage now decide access.For builders the rule is simple: don't build a product whose core value depends on an unlicensed path into someone else's platform. Get it in writing, keep a fallback, and assume today's open door has a toll booth planned.