Account settings

Your account page is intentionally compact. AgentSwarms is a learning and building platform, not a billing console, so we have kept the account surface to the things you actually need to change from time to time: who you are, how you sign in, what models you have connected, how much you are spending, and what we are allowed to email you about. Everything is collected on a single page at /account and the rest of this section walks through each block in detail.

Profile and identity

The profile block at the top of the account page is the public face of your work on AgentSwarms. Your display name and avatar are what appear next to anything you publish to the community marketplace, next to comments you leave on other people's agents, and on the certificate you receive when you complete the certification exam. We do not show your email publicly anywhere on the platform; only your display name is visible to other users.

• Display name can be changed at any time and is the only name shown publicly. If you change it later, all of your existing published agents and comments update to reflect the new name immediately — there is no orphaned history under your old name.
• Avatar can be uploaded directly (PNG or JPG up to a few megabytes) or pulled in from your Google account if you sign in with Google. If you upload nothing, we render a deterministic generated avatar from your user ID so your profile never looks empty in the community feed.
• Public bio is an optional short paragraph that appears on your community profile page, giving other learners context about who built the agent they just liked.

Authentication and sign-in

Authentication runs through the same secure backend that powers the rest of the platform. You can change your email address, change your password, and connect or disconnect Google sign-in entirely from the account page; you should never need to leave the app for any of these flows.

• Change email sends a verification email to the new address and a notification email to the old one. The change only takes effect after you click the verification link, which means a stolen session cannot silently move your account to an attacker- controlled address.
• Change password requires your current password as a confirmation step. After a successful change, every other existing session is signed out automatically as a safety measure, so a forgotten laptop in a coffee shop stops being a threat the moment you rotate your password.
• Google sign-in can be linked to an existing email/password account or used standalone. If you sign up with Google, you can later set a password to enable email/password as a backup sign-in method without disconnecting Google.

Subscription, credits and budgets

AgentSwarms is free forever for learning at a reasonable volume. Paid tiers exist for power users who run larger workloads, want bigger knowledge bases, need higher concurrency on swarm runs, or want the included certification voucher and faster support response times. The subscription block on your account page shows your current plan, the credit balance for usage-based features, and a running view of this month's spend across providers.

• Plan selection lets you upgrade or downgrade at any time. Downgrades take effect at the end of the current billing period, so you do not lose access to features you have already paid for that month. Upgrades take effect immediately and we prorate the difference.
• Credit top-ups apply to usage-based features (model inference through the AgentSwarms AI gateway, hosted embeddings, and image generation). Credits never expire and roll over indefinitely so you can pre-buy in bulk without worrying about losing unused balance.
• Spending controls let you set both a monthly cap per provider and a per-agent budget. When a cap is approached we send an alert email at eighty percent of the limit, and when the cap is hit we pause new runs rather than failing them silently, so you always have a chance to react before a runaway loop becomes expensive.

Bring-your-own provider keys

AgentSwarms supports a long list of model providers out of the box, and the account page is where you connect them. Once a provider is connected, every agent, swarm, notebook, and playground session can select its models from the list. Keys are encrypted at rest with a per-workspace key and are only ever decrypted server-side inside the backend functions that actually call the provider — they are never sent back to the browser, never logged, and never exposed in exports.

• Supported providers currently include OpenAI, Anthropic, Google Gemini, Grok (xAI), Azure OpenAI, AWS Bedrock, Google Vertex AI, OCI Generative AI, Alibaba Qwen, custom Ollama endpoints, and self-hosted vLLM endpoints. Theintegrations page covers the full list in depth.
• Health checks run automatically a few times a day on every connected provider. If a key is revoked, expires, or hits a quota wall, the integrations badge on the account page turns amber and any agent attempting to use that provider falls back to its configured backup model rather than failing.
• Removing a key immediately invalidates it server-side. Any in-flight requests using that key complete normally; any subsequent request from any agent that was using that provider falls back to the AgentSwarms AI gateway or to the agent's configured fallback model.

Data export and deletion

Everything you create on AgentSwarms belongs to you and we make it easy to take with you or remove entirely. Both controls live at the bottom of the account page under a clearly labelled "Your data" section, and both produce results within minutes rather than requiring you to email support.

• Data export generates a single downloadable archive containing every agent, swarm, skill, prompt, knowledge base manifest, notebook fork, trace, and community post tied to your account. Agents and swarms inside the archive are in a portable JSON format that can be re-imported into AgentSwarms or parsed by your own tooling.
• Account deletion initiates a thirty-day soft- delete window during which signing back in cancels the deletion. After thirty days the deletion is irreversible and removes all personal data, including community posts, comments, and the contents of your knowledge bases. Public traces of your contributions, if any, are anonymised rather than removed so as not to break other users' workflows that depended on your published agents.

Notifications

The notification block is finer-grained than the typical "all on / all off" toggle most platforms ship with. Each category of email and in-product notification can be turned on or off independently so you can stay informed about the things you care about without being drowned by the rest.

• Budget and usage alerts notify you when monthly provider spend crosses a configured threshold, when an individual agent's budget cap is approached, or when a knowledge base is near its storage quota.
• Approvals are the human-in-the-loop notifications that fire when a swarm or agent is waiting for you to approve a tool call before it executes. These are delivered both in-product and by email, and you can control which channels are active per category.
• Community activity covers comments on agents and swarms you have published, replies to comments you have left, and new likes on your top-performing published items.
• Certification updates include reminders about unfinished modules, exam scheduling confirmations, results, and delivery of the verifiable certificate after a passing attempt.
• Product announcements are kept deliberately rare and only sent when something materially changes about how the platform works (a new provider supported, a new template category, an updated certification rubric).