Coding & agents

Catch: 25 message credits a month runs out fast if you change your app often. Connecting your own domain and in-app code edits start on Starter ($16/month billed yearly).

Catch: Bolt branding on websites; 10 MB upload limit.

Catch: Chat and agent use draw on a small GitHub AI Credits allowance. More use needs Pro ($10/month); frontier models need Pro+ ($39/month) or Max ($100/month).

Catch: Basic weekly rate limits. Claude Sonnet 4.6, Claude Opus 4.6 and GPT-OSS-120b leave the Free tier on November 2, 2026; Gemini 3.8 Flash stays.

Catch: 3 credits per 30 days is a small allowance, and some restrictions apply depending on the IDE.

Catch: Newer models (starting with Gemini 3 Pro) and higher task limits come with Google AI Pro or Ultra.

Catch: It uses one smaller model that JetBrains can swap without notice, so hard multi-file work still needs paid frontier models. Requests stop if the model hits capacity, and the launch post says the offer runs until December 31, 2026.

Catch: Claude Sonnet 4.5 on Free requires signing up with a social login or AWS Builder ID, and usage is rate-limited.

Catch: Daily credit cap; credits are not refundable.

Catch: Free usage limits aren't published and access is subject to rollout; Codex on the web, in the CLI and IDE extension and on iOS starts with Plus ($20/month).

Catch: Free models are offered for a limited time; using OpenCode's pay-as-you-go models means adding billing details.

Catch: Limits for the free key are not published; paid access with a larger context window goes through OpenRouter.

Catch: The free published app goes offline after 30 days and carries a "Made with Replit" badge; full builds need Core ($20/month).

Catch: The $5 credit is one-time, not monthly, and Free is capped at 7 messages a day.

Catch: windsurf.com/pricing now redirects to devin.ai/pricing, where the editor is sold as Devin Desktop. The quota size isn't published, and frontier models need Pro ($20/month).

Catch: Zed-hosted models and unlimited edit predictions are on Pro ($10/month); on Free, AI runs on your own keys (billed by your provider) or local models.

Catch: No hosted plan; you pay your own model API costs unless you run a local model.

Catch: Only 100 conversations, with no top-ups on the Free plan.

Catch: Free software, but you pay your own model inference costs.

Catch: Capped at 50 workflow executions a month; enterprise connectors are not included.

Catch: Exact usage limits are not published; cloud agents, automations and AI code reviews are paid-plan only.

Catch: The 200 message credits are for trying it out; after that you use your own API key. The self-hosted Community edition must comply with the Dify Open Source License.

Catch: The daily refill stops for good once an unpublished lifetime allowance of free credits is used up.

Catch: The free usage amount is not stated; paid plans ($20 and $100 a month) add 2x and 10x usage, with no extra features.

Catch: Model usage is billed separately.

Catch: The Free plan includes no AI chats, so it runs scripted flows only. AI chats start on Starter (€32/month billed annually, 100 AI chats).

Catch: Free cloud limits are not stated by vendor.

Catch: Manus's pricing page now shows only paid plans (the Free plan is described in its help center); Pro starts at $20/month.

Catch: Single user only, and you pay your model provider. The pricing page no longer lists a free hosted cloud plan; multi-user Teams is $499/month for up to 10 users.

Catch: The credits are a one-time pool, not a monthly allowance, and the page does not say how many credits each action uses.

Catch: Compute and marketplace tools are billed from the run balance; no API or CLI access on Free.

Catch: You run and pay for hosting and inference; the hosted Cloud version is a 7-day trial only.

Catch: The credits are a one-time allowance; continued use needs a paid plan (Team is $100 a month).

Catch: Publishing a native app to the App Store or Google Play starts on Builder ($65/month; first month $30).

Catch: Only 10 rules per index, a 1 GB index limit and 7 days of analytics.

Catch: Logs are kept 1 day, MFA beyond basics and role management are paid, and a custom domain needs card verification.

Catch: Data is kept for 30 days, support is community only, and Personal is meant for individual developers and small projects.

Catch: Logs and traces are kept only 3 days, and on-call responders are paid seats.

Catch: Data is kept 14 days; extra processed data is $4/GB and extra scores $2.50 per 1k.

Catch: Free apps stay in the development version. A live website, custom domain and app store publishing need a paid plan (Starter from $32/month billed annually).

Catch: The Hobby plan is hard-capped at 10k API and 1k browser check runs a month; more needs the Starter or Team plan.

Catch: The official page doesn't list a catch.

Catch: Everything you make is public, and the free plan shows ads. Private Pens, CodePen AI and asset hosting start on Starter ($8/month billed yearly).

Catch: The Free plan is for personal development; Forge Inference and Sandboxes are billed monthly, with a $5 a month inference credit listed only for a limited time.

Catch: Deepnote AI is tiny on Free: 10 code completions and 5 agent calls a month. Machines shut down after 15 minutes idle, and scheduled runs and background execution need Team ($39 per editor/month).

Catch: Hard limits with no overages: you must upgrade to go past them, and a custom domain starts on Pro ($249 a month, billed annually).

Catch: The official page doesn't list a catch.

Catch: Free builds wait in a low-priority queue and time out after 45 minutes. Allowances reset each month and do not roll over; Starter is $19/month plus usage.

Catch: Low rate limits.

Catch: Only 5 AI generation requests for the life of the account. Code download, APK download and custom domains start on Basic ($39/month).

Catch: Past 7,500 users or 5 enterprise connections, usage is billed pay as you go.

Catch: Until January 1, 2027, selling a game needs the one-time Professional licence. Free commercial use covers only GMRT, which is still in beta on some platforms, and console exports always need an Enterprise licence.

Catch: Included hours count against a usage multiplier tied to machine size (2 for a 2 core machine), and without a payment method on file usage is blocked once the quota is used up.

Catch: Credits are limited and publishing apps or enabling workflows starts with the $25/month Basic plan.

Catch: Free GPU access is heavily restricted and not guaranteed, notebooks run for at most 12 hours, and execution stops when you stop interacting; faster GPUs need compute units (Pay As You Go, Colab Pro or Pro+).

Catch: Data is kept 14 days, only 3 active users, and support is community only.

Catch: Logs are kept only 7 days, with one seat and low ingestion and API rate limits.

Catch: Free GPU and TPU availability is limited, so you may wait in a queue at busy times; sessions end after 20 minutes idle.

Catch: Only 2 custom roles, 10 feature flags and 1 webhook; if you bill customers through Kinde it charges 0.7% per transaction.

Catch: Only 30 days of data access.

Catch: Only 1 seat, base traces are kept 14 days, and usage past 5k traces is billed pay as you go.

Catch: GPU time is pay as you go once the credits are used, and the free Studio restarts every 4 hours.

Catch: Only 50 AI credits a month.

Catch: Ingestion pauses at the 10M cap, and there is one seat plus two read-only guests.

Catch: Logs are kept for 7 days and ingestion is capped at 412 logs a minute; the page does not say whether the log and score allowances reset monthly.

Catch: AI credits don't roll over, and external users, embedded apps, Git source control and SSO aren't included.

Catch: Scanners run at view depth only on Free; full scan depth starts on the $9 a month Starter plan. API calls past 10,000 a month cost $0.10 per 1,000.

Catch: Limited to one user.

Catch: File uploads are capped at 1 MB per project, and you can't connect to APIs on localhost or CORS-protected APIs. Private team repos need Teams ($55 per member/month billed yearly).

Catch: It is a pay-as-you-go plan: usage past the free allowance is billed (extra SSO or SCIM connections are $125 each), and removing Stytch branding is a $99 add-on.

Catch: UptimeRobot pitches Free for hobby and non-profit projects; checks run only every 5 minutes, with 5 integrations and no team seats.

Catch: Free AI usage amount is not given.

Catch: Enterprise SSO and Directory Sync are billed per connection ($125 a month each for the first 15), and each extra million users is $2,500 a month.