Question 1

Sign up for SAMI

Accepted Answer

Create your account with an email address or Microsoft login in under two minutes. Creating a SAMI account takes under two minutes. Go to sami-agent.com and click "Get started". You can register with any email address or continue with your Microsoft account via OAuth — no separate password needed for the OAuth path. After registering with email, check your inbox for a verification message and click the confirmation link. Your account is active immediately after verification. The Free plan is available from day one with no credit card required. If you want to upgrade to Pro, Team, Ultra, or Enterprise, you can do so from the Billing section of your dashboard at any time.

Question 2

Install the desktop app

Accepted Answer

Download and install the SAMI desktop app on Windows, macOS, or Linux. The SAMI desktop app is where all AI work happens — chat, multi-agent consensus, file editing, and terminal. Visit sami-agent.com/download and select your platform: Windows (x64 or ARM64), macOS (x64 or Apple Silicon), or Linux (x64 or ARM64). Run the installer. On Windows, use the standard installer for a user-level install (no admin rights required). On macOS, open the DMG and drag SAMI to your Applications folder. On Linux, use the provided .deb or .AppImage. Sign in with your SAMI account after installation. The app connects to your account and syncs your plan, credits, and settings automatically. You do not need to configure any API keys — SAMI manages model access through your subscription.

Question 3

Your first consensus run

Accepted Answer

Learn how to kick off your first AI council session and read the results. Open the SAMI desktop app and navigate to the War Room. Type your task in the input field — for example, "Review this function for edge cases" and attach a file or paste code. Press Enter or click "Run". SAMI's AI council kicks off immediately. You will see each stage of the pipeline appear in real time: the council analyses your task, reaches consensus, and delivers a synthesised result. The whole process is transparent — you can read each AI's reasoning, see where they agreed, and where a veto was cast. If the council needs to clarify something before proceeding, it will ask you a short question inline. Answer it and the pipeline continues. Once complete, the final result appears at the bottom of the War Room view. Your first run uses credits from your plan. Free-tier users get a monthly allocation to try the full experience.

Question 4

Choose your verbosity mode

Accepted Answer

Control how much detail the AI council shows while it works. SAMI offers three verbosity levels you can set per session or globally in Settings. "Quiet" shows only the final result — useful when you just want the answer. "Normal" shows stage-by-stage progress indicators. "Verbose" shows each AI's individual reasoning, vote, and the consensus logic. To change the mode, open Settings in the desktop app and look for "Response verbosity" under the AI section. You can also toggle it inline from the War Room toolbar before starting a run. Quiet mode is the fastest for experienced users who know what to expect. Verbose mode is helpful when debugging or when you want to audit the council's reasoning before trusting the output.

Question 5

What are the 7 pipeline stages?

Accepted Answer

A plain-English walkthrough of each stage in the SAMI AI pipeline. SAMI runs your task through a structured pipeline before delivering a result. Think of it as an assembly line where each station has a specific job. Stage 01 (Connector) gathers the context it needs. If anything is ambiguous, it asks you targeted clarifying questions inline — for as many rounds as it takes to remove the ambiguity, bounded only by your run budget — before the pipeline continues. Stage 02 (Strategist) analyses the problem space and plans the approach. Stage 03 (Architect) locks the system design and interface contracts. Stage 04 (Maverick) is where one council member writes the binding code draft while the rest critique and challenge. Stage 05 (Optimizer) refines and stress-tests the draft. Stage 06 (Guardian) is a red-team and security audit — all council members check for vulnerabilities, false assumptions, and missed edge cases. Stage 07 (Debugger) is the final validation pass; only a blocking vote can hold output here. Not every run uses all seven stages — the pipeline adapts based on task complexity. A simple question may skip intermediate stages. A large refactoring job uses all seven and may iterate within stages. You can watch each stage in real time from the War Room panel. The pipeline bar at the top shows which stage is active and which are complete.

Question 6

What is a veto round?

Accepted Answer

How the council casts vetos and what happens when AIs disagree. A veto happens when a council member votes no on the proposed output. The member must explain the objection, and the council resolves it before work continues. Rather than silently overriding each other, the council surfaces the disagreement so it can be resolved before the result reaches you. When a no vote is cast, the pipeline pauses at that stage. The council works through the objection — either by refining the output or by escalating to you if resolution requires more context. This prevents the council from shipping a result that has a known flaw. In practice, no votes are rare. They surface mostly on tasks with ambiguous requirements, security-sensitive operations, or when the requested action conflicts with known constraints. You will see a clear "Veto" indicator in the War Room alongside the reasoning. Every objection is treated the same way — raised as a no with a reason, then resolved by the council. There is no sensitivity dial to tune.

Question 7

Reading the council's debate

Accepted Answer

How to interpret the AI council transcript in the War Room. The War Room shows you the council's internal work as it happens. Every selected model participates as a full council member with the same voting weight. For each stage, one member coordinates synthesis and one records the binding output; the rest challenge and vote on equal footing. You do not need to track which AI model has which extra duty — SAMI assigns that automatically. Look for the coloured badges next to each message: green means the model endorsed the step with a yes vote, red means it cast a no — a veto it must explain in writing. Voting is binary, so there is no in-between "maybe" badge; a model with any doubt votes no and states why. The final synthesised answer appears at the bottom of the transcript with a summary of how the council resolved every no into a genuine yes. If you are reading a completed run (from the replay view), you can click any stage to expand the full reasoning for that step. This is useful for auditing or understanding why the council made a specific choice.

Question 8

When the AIs disagree

Accepted Answer

What to do when the council cannot reach consensus automatically. Most tasks resolve to consensus automatically. When they do not, SAMI escalates to you in one of two ways: either asking a clarifying question inline (which you answer in the chat), or surfacing a "Review needed" state where you choose between two competing outputs. If the council is stuck on an architectural decision — for example, two equally valid approaches with different trade-offs — it presents both options with a summary of the trade-offs. You pick one and the pipeline continues from that point. You can also reduce escalations by providing more context upfront. A clear task description with explicit constraints ("prefer performance over readability", "must be backward-compatible with Python 3.10") gives the council enough signal to resolve disagreements without asking you.

Question 9

Compare SAMI plans

Accepted Answer

Side-by-side overview of Free, Pro, Team, Ultra, and Enterprise plans. SAMI has five plans. Free ($0/mo) gives you a monthly credit allocation, access to the managed model catalog, and full RAG functionality. It is a real working plan, not a trial — suitable for individuals exploring the tool. Pro ($50/mo) increases your credit allocation significantly and removes most usage throttles. Team ($150/seat/mo) adds shared team budgets, seat management, and SSO. Ultra ($400/mo) is for power users who need the highest credit limits and priority queuing. Enterprise is custom-priced and adds bring-your-own-key (BYOK), DPA/BAA documents, and dedicated support. All paid plans include full RAG access, the complete model catalog, and all pipeline features. The difference between tiers is primarily credit allocation and organisational controls. Visit sami-agent.com/pricing for current plan details — prices are served live from the backend and reflect any promotions.

Question 10

What happens when I hit my quota?

Accepted Answer

How SAMI handles credit exhaustion and your options for continuing. When your monthly credit allocation runs out, SAMI stops processing new AI runs until the quota resets. You will see a clear "Quota reached" indicator in the desktop app and dashboard. Existing runs that are mid-pipeline are allowed to finish. Your quota resets at the start of your next billing cycle. You do not lose any data, settings, or history — only new runs are paused. If you need to continue before the reset, you have two options: top up with additional credits from the Billing section of your dashboard, or upgrade to a higher plan. Upgrades take effect immediately and are prorated for the remaining billing period.

Question 11

Refund policy

Accepted Answer

When and how you can request a refund for your SAMI subscription. SAMI offers refunds for subscription payments made in error or for plans that were not used. If you subscribed and realised the plan is not right for you within 7 days, contact support@sami-agent.com and we will process a full refund. Credits that have been consumed (used for AI runs) are not refundable. Unused credits within a cancelled subscription are not automatically refunded, but we review requests case by case — reach out to support with your account details. For billing disputes, you can also contact your card issuer. If you believe a charge is incorrect, email us first — we can usually resolve it faster than a chargeback. Refunds are processed within 5-10 business days and returned to the original payment method via Stripe.

Question 12

Where does my code and chat content live?

Accepted Answer

SAMI processes your input to generate a response, but does not store your chat content server-side. SAMI is designed around a local-first privacy model. When you send a message or attach a file in the desktop app, the content is sent to SAMI's servers to process the AI run and stream back a response. Once the response is complete, your chat content is not retained on SAMI's servers. Your conversation history is stored locally on your device by the desktop app. SAMI does not have server-side access to your chat history. If you uninstall the app or clear local data, that history is gone — we cannot recover it. Usage metadata (credit consumption, run counts, timing) is stored server-side to power your dashboard analytics. This data does not include the content of your messages or files.

Question 13

GDPR Art. 15 — your right to access your data

Accepted Answer

How to request a copy of the personal data SAMI holds about you. Under Article 15 of the GDPR (and its German equivalent, the DSGVO), you have the right to request a copy of the personal data SAMI holds about you. This includes your account details (name, email, plan history), usage statistics, and any profile settings. To submit a data access request, email privacy@sami-agent.com with the subject line "Data Access Request" from the email address associated with your account. We will respond within 30 days with a structured export of your data. Remember that SAMI does not store your chat content server-side — the personal data we hold is limited to account and usage metadata. The export will reflect this scope.

Question 14

Delete your account

Accepted Answer

How to permanently delete your SAMI account and all associated data. To delete your account, go to Dashboard > Settings > Account and click "Delete account". You will be asked to confirm with your password. Deletion is permanent and irreversible. When you delete your account, SAMI removes your email, profile, plan history, and all server-side usage records within 30 days. Billing records are retained for the legally required minimum period for tax purposes, as required under German and EU accounting law. Your local chat history (stored on your device) is not affected by account deletion — you need to clear that separately from the desktop app. After deletion, your subscription is cancelled immediately with no further charges.

Question 15

Bring your own keys (BYOK)

Accepted Answer

Enterprise customers can supply their own AI provider credentials for maximum data control. BYOK (Bring Your Own Key) allows Enterprise customers to supply their own AI provider API credentials. When BYOK is active, AI inference requests are routed through your own provider account — SAMI acts as the orchestration layer, but your data flows under your API contract with the provider. This is the strongest data-isolation option and is designed for regulated industries where data residency or provider selection is a compliance requirement. It requires an Enterprise plan and is configured by your account administrator. Free, Pro, Team, and Ultra plans use SAMI's managed model catalog exclusively. BYOK is not available on these tiers. If you need BYOK, contact sales@sami-agent.com to discuss your requirements.

Question 16

How does SAMI pick which AIs to use?

Accepted Answer

SAMI selects models automatically based on task type, cost, and performance signals. SAMI manages a curated catalog of AI models. When you start a run, SAMI's routing layer analyses the task — its complexity, domain, and the stage of the pipeline — and selects the most appropriate model for each step automatically. You do not need to choose a model manually. The goal is to give you the best output at the lowest cost for your plan tier. More complex tasks at higher-tier plans get access to larger, more capable models. Simpler tasks use lighter, faster models to preserve your credits. The catalog is updated regularly. When a provider retires a model, SAMI substitutes it automatically — you will not hit a "model not found" error mid-run. If you are on an Enterprise plan with BYOK, you can additionally select specific provider models through the admin panel.

Question 17

Why a council instead of one model?

Accepted Answer

Multiple AI perspectives catching each other's mistakes before the result reaches you. A single AI model can be confidently wrong. The council approach runs multiple independent AI agents through a structured pipeline where each one can challenge the others. The result you receive has survived internal review — not just generation. Think of it like peer review. A single expert writing a report can miss edge cases or make assumptions. Four experts reviewing each other's work catch more issues before the paper ships. SAMI's council applies the same principle to AI output. The trade-off is speed — a council run takes longer than a single-model call. SAMI mitigates this with a structured pipeline that runs stages in sequence but makes each stage as efficient as possible. For tasks where correctness matters more than raw speed, the council is the right tool.

Question 18

Free tier vs paid tier: which models do you get?

Accepted Answer

Free users access a capable subset of the catalog; paid tiers unlock the full range. All SAMI users — including Free — access the managed model catalog. The difference between tiers is which models in that catalog are routed to your runs, and how many credits you have available each month. Free tier runs are routed to efficient, capable models well-suited for everyday tasks. Paid tiers unlock routing to larger, higher-capacity models that handle more complex tasks, longer contexts, and specialised domains. Ultra tier gets priority access to the most capable models in the catalog. Model names and specifics are intentionally not published on public pages — the catalog evolves as providers release new models, and locking to specific names would make every policy update stale. What you can rely on: the routing optimises for your task at every tier.

Question 19

Add team members

Accepted Answer

Invite colleagues to your SAMI team and manage their access. Team and higher plans include seat-based team management. To add a member, go to Dashboard > Team and click "Invite member". Enter their email address and select a role. They will receive an invitation email with a link to create their account and join your team. Each seat you add is billed at the per-seat rate for your plan. The invited user inherits your team's credit pool and model access immediately upon accepting the invitation. Team admins can remove members, change roles, and manage the shared budget from the same Team page. Members on a team cannot exceed the team's monthly credit limit — if the pool is exhausted, all members are paused until the next billing cycle or until credits are topped up.

Question 20

Shared budget across your team

Accepted Answer

How the team credit pool works and how to set per-member spending limits. On Team plans, credits are pooled across all members by default. Any team member can draw from the pool, and the balance is visible to all team admins on the Team dashboard. You can optionally set per-member spending limits to prevent any single member from consuming a disproportionate share of the monthly budget. Go to Dashboard > Team, click a member's name, and set a monthly cap in the "Spending limit" field. Members who hit their cap are paused without affecting other team members. The team admin (the account that created the team) always has full visibility into credit consumption broken down by member. This report is available from Dashboard > Analytics.

Question 21

SAMI's compliance posture (honest status)

Accepted Answer

Where SAMI stands today on SOC 2, ISO 27001, and other certifications. We believe you deserve an honest answer, not a marketing one. As of today, SAMI does not hold SOC 2 Type II or ISO 27001 certifications. These are on the roadmap and we are building toward them — but we will not claim them before the audit is complete and the certificate is in hand. What we do have: end-to-end transport encryption, server-side encryption at rest, access controls with audit logging, and a local-first data model that minimises what we store about your work. Our Privacy Policy accurately reflects what data we hold and for how long. If your organisation requires specific certifications before onboarding, Enterprise is the appropriate tier — contact sales@sami-agent.com to discuss your requirements. We can provide a security questionnaire response and engage with your InfoSec team directly.

Question 22

Custom provider credentials for regulated industries

Accepted Answer

Enterprise customers can route AI inference through their own provider accounts. Regulated industries (healthcare, finance, legal, government) often require that AI inference does not cross certain data boundaries. SAMI's Enterprise tier addresses this with BYOK (Bring Your Own Key): your AI credentials, your provider contract, your data agreement. With BYOK configured, SAMI orchestrates the pipeline but all model API calls go through your organisation's provider account. This means your data is subject to your existing data processing agreement with that provider — SAMI never sees the inference payload at the provider level. To configure BYOK, an Enterprise admin navigates to Admin > Providers and enters the credentials. Multiple providers can be registered and the routing layer selects the appropriate one per task. Contact sales@sami-agent.com to get started.

Question 23

Connect an external AI assistant

Accepted Answer

Use SAMI integration tokens to let an approved external assistant access audited capabilities. SAMI can expose an audited integration bridge for approved external AI assistants, agent frameworks, or custom integrations. This is different from the desktop app: the integration path is for programmatic, administrator-controlled access. To connect, an administrator generates a scoped integration token from the admin panel. The token grants access only to the capabilities selected for that integration. The external assistant includes this token in its requests to SAMI's integration endpoint. Integration tokens are scoped and can be revoked at any time. Do not share them or commit them to source control. Each external integration should have its own token so you can revoke one without affecting others.

Question 24

Managing integration tokens

Accepted Answer

Create, rotate, and revoke external integration tokens from the admin panel. Integration tokens are managed from the admin panel. Each token has a name so you can identify which integration uses it, a creation date, and an optional expiry. There is no limit on the number of tokens you can create. To revoke a token, click the trash icon next to it in the token list. Revocation is immediate - any request using that token will be rejected after revocation. Create a new token for the integration before revoking the old one to avoid downtime. Best practice: give each external system its own token, name it clearly ("Staging assistant"), and set an expiry date for tokens used in automated pipelines. Treat integration tokens like API keys - rotate them periodically and revoke any that may have been exposed.

Question 25

Platform support

Accepted Answer

Which operating systems and architectures the SAMI desktop app supports. The SAMI desktop app runs on Windows (x64 and ARM64), macOS (x64 and Apple Silicon), and Linux (x64 and ARM64). All platforms receive the same features — there is no "primary" platform. Minimum requirements: Windows 10 or later, macOS 12 or later, or a modern Linux distribution with a glibc >= 2.31. RAM: 4 GB minimum, 8 GB recommended for large file contexts. The app stores local data in your user profile directory — no admin rights required on Windows. There is currently no mobile (iOS / Android) app. The web dashboard at sami-agent.com is mobile-responsive for account management tasks, but the AI work surface requires the desktop app.

Question 26

How SAMI updates itself

Accepted Answer

The desktop app checks for updates automatically and applies them in the background. The SAMI desktop app checks for updates automatically when it is running. When an update is available, it downloads in the background while you work. You are prompted to restart the app to apply the update — no forced interruption. If you prefer to control when updates apply, you can defer the restart. The app will remind you periodically until the update is applied. Security updates are applied on the next restart regardless of deferral preference. You can always check your current version from Help > About in the app menu. The changelog for each release is available at sami-agent.com/changelog.

Question 27

Enable two-factor authentication (2FA)

Accepted Answer

Add a second layer of security to your SAMI account with a TOTP authenticator app. Two-factor authentication (2FA) adds a second verification step when you sign in. SAMI uses time-based one-time passwords (TOTP), compatible with apps like Google Authenticator, Authy, or 1Password. To enable 2FA, go to Dashboard > Settings > Security and click "Enable two-factor authentication". Scan the QR code with your authenticator app and enter the 6-digit code to confirm setup. You will also be shown backup codes — save these in a safe place. If you lose your authenticator device, backup codes are the only way to recover access. Once enabled, every sign-in requires your password plus a TOTP code. 2FA cannot be bypassed remotely by support staff — if you lose both your authenticator and your backup codes, account recovery requires identity verification and takes up to 5 business days.

Question 28

Change your username or email address

Accepted Answer

Update your display name and email from the Settings page. To change your display name, go to Dashboard > Settings > Profile. Edit the "Name" field and click Save. Changes take effect immediately and are reflected in your dashboard and any shared team views. Changing your email address requires verification. Go to Dashboard > Settings > Profile, click "Change email", enter your new address, and click Send verification. A link is sent to the new address — click it to confirm the change. Your old email continues to work until you confirm the new one. If you signed up via Microsoft OAuth, your display name is pulled from your Microsoft account. You can override it in the Profile section of the dashboard.

Question 29

Email notification settings

Accepted Answer

Choose which emails SAMI sends and opt out of the ones you don't need. SAMI sends three categories of email: transactional (account verification, password reset, billing receipts — these cannot be disabled), security (new device login alerts, 2FA changes), and product updates (release notes, usage summaries). To manage your notification preferences, go to Dashboard > Settings > Notifications. You can toggle product update emails and security alert emails independently. Transactional emails are always sent because they are required for account function and legal compliance. To unsubscribe from marketing emails, use the unsubscribe link at the bottom of any product update email. Alternatively, toggle the preference off in Settings. Unsubscribing takes effect within 24 hours.

Question 30

Login fails — what to check

Accepted Answer

Step-by-step checklist for diagnosing sign-in failures on the web and desktop app. If your login is failing, start with the basics: confirm you are using the correct email address and that Caps Lock is off. SAMI passwords are case-sensitive. If you get an "Account not verified" error, check your inbox (and spam folder) for the original verification email and click the link. If it has expired, use the "Resend verification" link on the login page. If the password is correct but login still fails, try resetting your password via the "Forgot password" link. If you signed up with Microsoft OAuth, click the "Continue with Microsoft" button instead — you cannot use a password if OAuth was your original sign-in method. If you are locked out after multiple failed attempts, the lockout lifts automatically after 15 minutes. For persistent issues, email support@sami-agent.com with your account email and we will investigate.

Question 31

Streaming hangs mid-run

Accepted Answer

What to do when an AI pipeline run stops responding before it finishes. If a run stops outputting content mid-way and the pipeline bar is frozen, first wait 30 seconds — some pipeline stages process internally before emitting output. The hang may resolve on its own. If it is still stuck after a minute, check your network connection. Streaming requires a stable connection; a brief dropout can cause the stream to stall. The desktop app will attempt to reconnect automatically, but if it cannot, the run enters an "interrupted" state. You can re-run the task from the War Room. If runs hang consistently, try restarting the desktop app. On rare occasions, a background process can accumulate stale state that causes new runs to stall. A restart clears this. If the issue persists across restarts, check sami-agent.com/status for any active incidents.

Question 32

A worker crashed — what now?

Accepted Answer

How to recover when an AI pipeline worker exits unexpectedly. Worker crashes are rare but can happen if a run encounters an unexpected input, a transient infrastructure hiccup, or a timeout on a very long operation. The desktop app will show the run in an "Interrupted" or "Failed" state. Credits for interrupted runs are typically not deducted — the billing system detects the abnormal exit and reverses the charge within a few minutes. Check your usage dashboard to confirm. If credits were deducted for a failed run, contact support@sami-agent.com with the run ID shown in the War Room. To recover, simply re-run the task. If the same run crashes repeatedly, try reducing the input size (shorter code, fewer attached files) to isolate whether the input is causing the issue. If you suspect a bug, use Help > Report a bug to send the run ID and error details to our team.

Question 33

Browser compatibility

Accepted Answer

Which browsers work with the SAMI web dashboard and known limitations. The SAMI web dashboard (sami-agent.com) is tested on the latest versions of Chrome, Firefox, Edge, and Safari. We recommend Chrome or Edge for the best experience — they have the most complete support for the web APIs the dashboard uses. Safari on iOS and macOS works but may have minor visual differences in some layouts. Firefox works fully. Internet Explorer is not supported. The AI work surface (chat, consensus runs, file editing) is only available in the desktop app — not in any browser. If a browser-based feature is not working, try clearing your browser cache, disabling extensions, or switching to an incognito window to rule out extension conflicts.

Question 34

Slow performance in the desktop app

Accepted Answer

How to diagnose and improve SAMI desktop app performance. If the SAMI desktop app feels slow, the most common cause is a large workspace with many files open. SAMI indexes your workspace for context — very large repositories (hundreds of thousands of files) can take time to index on first open. Subsequent opens are faster because the index is cached. Try closing file tabs you are not actively working with and removing large binary files from the workspace root. If you have more than 50,000 files in your workspace, consider creating a subfolder and opening that instead of the full repo. If the app is slow even with a small workspace, check your system's available RAM. SAMI needs at least 4 GB to run comfortably; 8 GB is recommended. Quit other memory-intensive applications if needed. A full app restart (not just closing and reopening the window) clears any accumulated state.

Question 35

How to report a bug

Accepted Answer

The fastest way to get a bug in front of the SAMI team. The fastest way to report a bug is from within the desktop app: go to Help > Report a bug. This automatically attaches diagnostic information (app version, OS, error logs from the last run) and opens a submission form. Describe what you expected to happen and what happened instead. If the bug is on the web dashboard, use the feedback button at the bottom of any dashboard page, or email support@sami-agent.com with "Bug report" in the subject line. Include your browser version, the page URL, and any error messages you see. Good bug reports include: steps to reproduce (numbered list), what you expected, what actually happened, and your app version (visible in Help > About). The more specific you are, the faster we can fix it.

Question 36

Beta program access

Accepted Answer

How to join the SAMI beta program and what it involves. The SAMI beta program gives participants early access to features that are still being tested and refined. Beta features may be incomplete, have rough edges, or change before general release. They are real features under active development — not demos. To join, email contact@sami-agent.com from the address associated with your SAMI account. Include a short note about what you are building with SAMI and which types of features you are most interested in testing. We select participants based on use-case fit — not first-come-first-served. Beta participants get access to a dedicated feedback channel and are asked to share structured feedback on the features they try. This is a two-way commitment: you get early access, we get high-quality feedback from real-world usage.

Help Center

Help Center