We ran four identical briefs on AutoAGI and AutoGPT: time, cost and results

How we ran it

• The identical text was pasted once into each product on September 29, 2026, with default settings and no follow-up messages while a task ran.
• AutoAGI ran on the Max membership ($200, $46 weekly allowance). AutoGPT ran on its 7-day Pro trial ($50 a month after the trial) in Otto chat. These are different tiers, so we compare behavior and consumption, not plan price.
• AutoGPT's separate automation-credit wallet held $3.00 and was never touched: all four tasks ran inside the Otto chat.
• One run per brief. AI output varies from run to run, so a second run could differ.

Results at a glance

Brief 1: a research brief a client could read

The brief: "Write a comparison brief on Lindy, Relevance AI and Gumloop for a 10-person marketing agency choosing an AI agent platform. For each: what it is, the entry paid price and how usage is metered, what it is best for, and its biggest weakness. End with a recommendation. Use current sources and give a link for every price. Deliver it as a document I can send to a client."

We checked the anchor facts against each vendor's own pricing page the same day. Both products got all of them right: Lindy Plus at $29.99 per user, Relevance AI Pro at $29 ($19 billed annually) with its free plan retired, and Gumloop Pro at $37 with 20,000 credits.

• AutoAGI linked a source for every price, added review links for each weakness, kept a source ledger, disclosed that some review sites sell competing products, and listed what it could not verify. It left two placeholders for the client name and a disclosure line, and said so. Its recommendation was to pilot Gumloop first.
• AutoGPT produced a finished 2-page PDF with vendor links that was ready to send. Its recommendation was Relevance AI's Team plan, with Gumloop added for custom pipelines.

The two disagree on the recommendation. That is a judgment call on a 10-seat agency, and both explained their reasoning.

Brief 2: score 30 leads against an ideal customer

The brief: score 30 invented leads from 0 to 100 against an ideal customer (a founder-led agency or consultancy, 5 to 20 employees, in the US, UK, Canada or Australia, already using AI tools, with a budget of $500 or more a month), rank all 30 with a one-line reason, and list the top 5. The lead list is synthetic, so no real company data was used.

As a consistency check we scored every lead by counting how many of the five criteria it meets. Both products chose to weight business type most heavily, which the brief allows, so this check is not the only right answer.

AutoGPT's ranking matched the simple check more closely. AutoAGI ranked two leads that fit every criterion except business type lower, because it treated type as the main filter and said so. AutoGPT's run started with equal weights, noticed a four-way tie, reweighted type and re-scored, which is part of why it took 23 minutes.

Brief 3: build a calculator

The brief: "Build a single-page ROI calculator for a marketing agency considering an AI agent platform that costs $100 a month. Inputs: number of people, hours per person per week spent on reporting, hourly cost, and the percent of that work the agent takes over. Outputs: hours saved a month, dollars saved a month, net savings after the $100 plan, and payback in days. Responsive, clean design. Deliver one downloadable HTML file."

• AutoGPT delivered in 2m39s with correct formulas. It described the file as having no dependencies, but the page loads Tailwind CSS from a public CDN, so it needs an internet connection to look right.
• AutoAGI took about 5 minutes and delivered a 20 KB self-contained file with no external requests. Before handing it over it ran the page in headless Chromium at four screen sizes, checked contrast, tap targets, keyboard order, print output and the math, and fixed two issues it found.

We did not judge visual design by eye here. If you care about that, open both files and decide for yourself.

Brief 4: a routine that runs every Monday

The brief: "Set this up to run every Monday at 7:00 AM: a one-page brief on any changes in the last week to pricing or plans for Lindy, Zapier Agents and AutoGPT, with a link to the source for each change, and a line saying "no change found" when there is none. Run it once now so I can see the first result."

• Both saved a weekly routine and produced a baseline first run. Neither could truthfully report a change, because there was nothing earlier to compare against, and both said so.
• AutoAGI's first run was longer, with a table of plans and prices, and it said plainly that it could not confirm any change in the last seven days. AutoGPT's was a short note with a source link per vendor.
• AutoGPT reported Zapier Agents Pro at $50 a month, which we could not confirm against Zapier's own page. Treat that figure as unverified.

What we take from it

• Speed depends on the job. AutoAGI was far faster on scoring and AutoGPT on the build. The research brief was close to a tie.
• Cost visibility differs. AutoAGI showed every task in dollars. AutoGPT showed a percentage of a daily and a weekly allowance, and four tasks used 29% of the day on the Pro trial.
• Accuracy was strong on both for facts we could check, with one unverified figure from AutoGPT and one placeholder-laden but candid brief from AutoAGI.
• AutoAGI did more self-checking, and it did not always land closest to the simple check. AutoGPT did not show any checking of its build.
• If you want to build and own agents, AutoGPT remains the better fit. See AutoAGI vs AutoGPT for the full comparison.

Questions

Sources

• AutoGPT pricing (agpt.co) (no page date shown)
• AutoGPT home (agpt.co)
• AutoGPT GitHub repository and license split (platform under Polyform Shield 1.0.0, classic under MIT)
• AutoGPT releases (platform-beta-v0.8.1 published 2026-09-24)
• AutoAGI pricing
• Lindy pricing
• Relevance AI pricing docs
• Gumloop credits docs