Fig
Summer 2020AcquiredFig adds visual apps, shortcuts, and autocomplete to your terminal.
Pressure-test this opportunity
Explore the risks and possibilities with a prompt for ChatGPT, Claude, or your agent.
Fig adds visual apps, shortcuts, and autocomplete to your terminal.
Explore the risks and possibilities with a prompt for ChatGPT, Claude, or your agent.
Fig was a developer-tool company founded by Brendan Falk and Matt Schrage and accepted into YC's Summer 2020 batch. It attached visual applications to existing terminals, beginning with IDE-style command-line autocomplete built from an open corpus of declarative specifications. Amazon announced on August 28, 2023 that it had acquired Fig's technology and that the team joined AWS.[1]
This was an acquisition success with a partial product sunset, not a distress story. The team shipped CodeWhisperer for command line within months, that product became part of Amazon Q Developer, and current AWS code retains direct Fig components. Independent Fig ended on September 1, 2024. Autocomplete survived, but Scripts, Dotfiles, Plugins, and Servers had no planned replacements. Deal terms and exact code contribution remain undisclosed.
Brendan Falk was Fig's CEO, while Matt Schrage did most of the early coding.[2] After several early-2020 pivots, they focused on a problem in their own work. Falk's concise description was that “the Terminal was a huge pain point for us.”[2]
The decision was not to build another terminal emulator. Fig would attach visual tools to terminals and IDE consoles developers already used, preserving shells, workflows, and user choice. Technical exploration began in late April 2020. The founders committed on May 15 and shipped an MVP to users about two weeks later.[2]
The first architecture paired a native Swift desktop application with small web applications rendered beside the terminal. This let the team move quickly while still controlling window position and local integration. The early deployment stack moved from Replit to Heroku, AWS, and S3 as the product matured.
The corpus preserves one exact founder quote above, not two independent verbatim founder passages suitable for reproduction. The second canonical quote requirement therefore remains an explicit evidence gap.
Fig's first product was autocomplete, but its thesis was broader. Fig.js exposed shell events, local command execution, UI positioning beside the terminal, and insertion of text into an active session. The goal was an application platform for the command line.[5]
Autocomplete used declarative JavaScript specifications. A contributor could describe a command and its arguments in roughly ten lines, then share that specification with every user. At general availability, Fig supported more than 200 command-line tools, and more than 120 people had contributed to the open corpus.[5]
The local terminal integration was technically difficult. The first MVP reconstructed input through a macOS event tap, which the team found brittle. Fig said keystrokes were never stored or transmitted and that autocomplete ran locally.[9] The replacement, figterm, inserted a pseudoterminal passthrough, reconstructed the screen, and marked cells as prompt, suggestion, output, or edit-buffer content. The team rewrote figterm from C to Rust with Tokio and shared Rust components with a cross-platform desktop rewrite.
By acquisition, Fig claimed five products, hundreds of thousands of users, 22,000 GitHub stars, 13,000 Discord members, and 400 open-source contributors.[1] These are company-reported community metrics, not independent usage audits.
The first AWS product retained contextual completions and inline documentation for more than 500 CLIs, then added natural-language-to-Bash translation.[6]
Fig addressed individual developers who wanted safer, faster command-line work and engineering teams that wanted shared terminal knowledge. Its mission explicitly covered beginners, advanced engineers, and collaborative teams.[5]
No reliable ARR, revenue, paid-seat count, gross margin, burn, or market share was found. YC's archived job page reports a five-person team, but the date and acquisition headcount are uncertain.[10] The reported community scale shows reach, not commercial conversion.
Warp competed through a modern terminal editor, completions, and natural-language command suggestions.[11] Fish's built-in completions and zsh-autosuggestions already had direct access to the edit buffer, avoiding some of Fig's integration complexity.[9]
Fig's distinction was terminal neutrality plus an open specification corpus. It preserved existing terminals and shells rather than requiring wholesale migration. The tradeoff was maintaining deep local integration across environments while monetizing a client whose most visible knowledge asset was community-built.
Fig planned to keep individual and open-source use free while charging per user for hosted team products.[5] By August 2022, Free, Pro, and Enterprise tiers existed. Enterprise added SSO, access controls, compliance, and support; paid team features included private completion specifications, dotfiles, environments, secrets, and planned shared workflows.[12]
TechCrunch reported a $2.2 million seed. General Catalyst lists Fig as a portfolio company backed since 2020, and Fig named General Catalyst and Mango Capital as backers.[4][13]
Acquisition terms were undisclosed. No price, consideration mix, valuation, earn-out, retention terms, founder proceeds, or precise close mechanics were found. Amazon's annual report aggregates Fig with other immaterial acquisitions and does not isolate its economics.[14]
Fig's open specification system produced its most visible distribution. More than 120 people had contributed by general availability; the company later reported 400 contributors, 22,000 stars, and 13,000 Discord members.[5][1] No named paying enterprise customer or customer-authored case study was found.
The open withfig/autocomplete repository remained active after the product sunset, preserving the specification corpus.[15] Current Amazon Q CLI code includes figterm, fig_desktop, and fig_input_method, proving direct technical lineage without revealing the percentage of current code inherited from Fig.[16]
Fig's acquisition succeeded at transferring team, core command-line experience, and technical components into AWS. The subsequent sunset was selective rather than total.
Fig chose existing terminals instead of a new emulator. That let users keep their shells and workflows, while the open specs made many CLIs useful quickly. The cost was complex platform integration, shown by the move from macOS event interception to a Rust pseudoterminal layer.
Community specifications created breadth that a five-person team could not author alone. Launch discussion also questioned whether a venture-backed proprietary client should rely on open community labor.[3] The corpus outlived the standalone product, showing that shared assets can survive even when the commercial client disappears.
Within months, Falk announced CodeWhisperer for command line. By April 2024 it entered Amazon Q Developer. The current AWS repository retains named Fig components and dual MIT and Apache licensing.[16] This is strong product-lineage evidence, though no source attributes Amazon Q revenue or adoption to Fig.
The migration exported local data, installed the AWS product, transferred settings, and removed Fig. Autocomplete survived. Fig explicitly said Scripts, Dotfiles, Plugins, and Servers had no planned replacements.[8] The app-platform thesis narrowed into a command-line assistant inside a larger product.
The counterargument is that sunset means failure. The evidence points elsewhere: Amazon acquired the technology and team, shipped a descendant product, and retained code lineage. What ended was independent packaging and several adjacent products, not the strategic value of the autocomplete asset.