Fig

today, we're launching Fig, the full funnel marketing agent. In a world of disconnected tools and AI slop, marketing needed a change. no waitlist — public beta is now live at figfunnels [dot] com comment "FIG" and Shane (our co-founder) will personally walk you through the platform. this release comes with 5 key features (thread):

@AiwithShoaib Let's go!!

@figfunnels Wow amazing

Introducing One Workflow by Fig: Run your entire marketing system in one place. Powered by Fig’s full-funnel context, build, launch, and optimize your ads, pages, and actions inside a single workflow.

@XMonetizationC_ @Aiwithkumail Overdue :)

Finally — an AI that doesn’t just generate ads, but owns the entire funnel in one unified workflow. This is significant. 🔥 Most AI marketing tools are fragmented: one for copy, one for creatives, one for landing pages, another for optimization. Fig just collapsed all of that into a single, context-aware system. One Workflow means full-funnel memory, seamless handoffs between stages, and true end-to-end optimization instead of disconnected experiments. It’s the difference between using AI as a fancy assistant and using it as an actual marketing operating system. In a noisy creator and DTC economy, the winners won’t be those with the best single asset — they’ll be those who can build, launch, test, and iterate complete funnels at speed with coherent strategy. Fig is building exactly that infrastructure. Clean interface, ambitious vision, and genuinely useful execution. This is how modern marketing teams (and smart solo operators) will move in 2026 and beyond. Well played."

@AiwithZoaina Thank you!!

@figfunnels Great workflow demo! One interface for full funnel sounds efficient.

@ElaraGrace_AI 1000%

@figfunnels No more jumping between endless tabs. Fig just makes everything click.

@kyronis_talks lets go

@figfunnels Most AI gives you random ads.
Fig gives you the entire damn funnel.
Big difference.

@Aiwithkumail had to be done

@figfunnels This actually a game changing tool

@kamal_razzak show up for your brand

The founder's conviction on camera is the one thing your competitor can't copy. They can copy your product. Your offer. Your funnel structure. They cannot copy your founder's story, told in their own words. That's the moat. Start filming.

@ech0_speaks Thank you :)

@figfunnels This looks insanely clean Finally an AI that understands the full marketing funnel instead of just spitting out random ads. One workflow to build, launch & optimize everything? Game changer. Loving this direction, Fig team!

@heyfionaai Thanks :)

@figfunnels Excellent Share

Introducing Ad Management by Fig: Run your ads without the chaos. Powered by Fig’s full-funnel context, manage your campaigns in one simple workflow, launch, test, and scale.

@MehtabKarta 👋

whats the current best bulk ads uploader for meta?

@heyfatema Thank you :) Let's set up a call

@figfunnels This is amazing

Introducing Personas by Fig: Build your customer before you market to them. Powered by Fig’s full-funnel context, create AI personas you can talk to, learn from, and use to write your ads, landing pages, and more.

@argon_delta s-tier sign off

there are only 7 visual hook formats every winning ad uses. memorize these. ur creative team is probably using 2. > 1. PATTERN INTERRUPT (PI) — visual that does NOT look like an ad. animal close-up, hand-drawn note, screenshot of a text msg. > 2. TEXT OVERLAY (TX) — bold text on a static background, no human. high CTR for skeptics. > 3. AVATAR INTRO (AI) — close-up of one person's face, talking direct to camera. trust signal. > 4. UGC PROXY (UG) — selfie angle, kitchen/bathroom, person holding product. recognition. > 5. SPLIT (SP) — before/after side by side. zero ambiguity about the transformation. > 6. BRACKET (BR) — circled / arrow / drawn-on screenshot pointing at the proof. > 7. SLICE-OF-LIFE (SL) — micro-moment from the sufferer's day. the cracked nail on the steering wheel. the bra strap dent. most ur ads sit in 2 of these 7. that means 5 entire audience pools are uncovered. each one lands in a different Andromeda cluster. ur TAM is limited. no scale, no rarri. print all 7. test 1 winner per format. signed from the temple of scale, argon

@danmartell This is exactly why we launched AI Personas

If your audience doesn’t feel called out, they won’t lean in.

@codyschneider figfunnels [.] com LETS GO

its always the same buy the .com just literally, always buy the .com there is 30 years of trust built into the .com you get to piggyback on that from day one just buy the .com

@DamiDina @drsnow well that was a joke. But we do actually have an integration with them. Meta has allowed developers to build direct integrations for a few years now. We just attached it our agent so that you can manage your full marketing funnel in one place. Let's setup a call

@figfunnels @drsnow meaning?

Good News: Meta MCP now allows you to integrate Claude to your ad accounts/reporting Bad News: it's only rolled out to ~10% of accounts right now. Would have been nice to know before wasting time trying to fix something that's not even available yet!

@codyschneider this + web builder = Fig

facebooks ads CLI shipped yesterday its a gift from god here's why Zero boilerplate code The raw Marketing API requires you to write auth handling, pagination logic, output formatting, and error handling every single time. The CLI packages all of that into single commands like `meta ads campaign list`. An agent just shells out — no SDK, no wrapper code. 2. Deterministic, single-purpose commands. Each CLI command does one thing. `meta ads insights get`, `meta ads campaign create`, `meta ads adset update`. An agent doesn't need to reason about which API endpoint to hit, what HTTP method to use, or how to construct the request body. It maps intent → command 1:1. 3. Structured output formats built in. The CLI ships with `json`, `table`, and `plain` (tab-separated) output modes. An agent can request JSON and parse it directly without ever dealing with the API's nested response objects, cursor-based pagination, or inconsistent field naming. 4. Dry-run / preview before spend. This is huge for agents. The CLI supports preview commands that show exactly what would be created without touching Meta's API. An agent can generate a spec, preview it, get human approval, and *then* execute — keeping the human in the loop before any money moves. 5. Spec files as the unit of work. Campaign launches are defined as JSON spec files. This means an agent can write a file, commit it to git, diff it against last week's launch, and hand it to a human for review. It turns ad creation into a code review workflow rather than a series of imperative API calls. 6. Standard exit codes for error handling. The CLI returns predictable exit codes — `0` for success, `3` for auth errors, `4` for API errors. An agent can branch on these without parsing error message strings or catching API-specific exceptions. Makes retry logic trivial. 7. `--no-input` and `--force` flags for unattended execution. The CLI was explicitly designed for CI/CD and automation. These flags suppress all interactive prompts, so an agent running in a headless environment (cron job, Claude Code, Cursor) never gets stuck waiting for user input. 8. Environment variable auth — no token juggling. Tokens, secrets, and account IDs go in env vars, not command arguments. An agent doesn't need to manage token refresh flows, handle OAuth callbacks, or keep credentials in memory. The runtime environment handles it. 9. Built-in rate limiting and retries. The Marketing API's rate limits are notoriously painful — you have to implement exponential backoff, track your app-level and ad-account-level usage tiers, etc. The CLI handles retries and rate limiting internally, so the agent just fires commands and gets results. 10. Drop-in agent skill files. The CLI ships with a skill file you can drop into Claude Code, Cursor, or any agent that can run shell commands. The agent reads creatives, writes copy, builds the spec, previews, and launches. Each command is single-purpose so the agent stays deterministic, and spec files give the agent a reviewable plan before any spend. but let's be real here's the cons - It shipped yesterday with docs pointing to Meta's internal domain, so you're an early adopter debugging in the dark. - Every CLI invocation spawns a new Python process, so high-volume loops across dozens of ad sets will be significantly slower than a persistent API session. - You lose granular control over niche targeting fields, beta features, and the hundreds of parameters the Marketing API exposes that the CLI hasn't wrapped yet. - Multi-tenant is clunky — switching between Jed's, Worthy's, and your own ad accounts via env vars is worse than just passing `act_{id}` on every API call. - No async events or webhooks, so your agent can poll but can't listen, which kills real-time optimization loops.

@DamiDina @drsnow only on weekends

@figfunnels @drsnow You know someone at meta?

@codyschneider or use Fig and simplify this whole process

how to facebook ads for saas scrape pain points of your target audience write UGC scripts based on those pain points scripts to AI UGC ads avatar must look like the person you're selling to take raw UGC and remove silence and add captions bulk upload ads to facebook via facebook ads api ads go into testing campaign winners get their own ad set with dedicated budget CPMr gets too high, turn off ads deploy this exact agent on graphed .com

@mes28io don't make a commercial

the new UGC meta is “proof with personality” not polished testimonials not founder demos not influencer praise proof screen recording real reaction specific use case one tiny imperfection a caption that feels like it was written by someone annoyed people don’t trust ads because ads are too smooth smooth feels planned planned feels paid paid feels suspicious the best creative feels like someone accidentally exposed a shortcut that’s why raw beats premium so often