LLM Model Selection (High Impact)
The strongest impact on your bill. Using Claude Sonnet or ChatGPT for basic support is a common mistake. Use the default Swiftask model for routine tasks.
High Impact
1 / 9
01 — How It Works
+
+
+
+
×
Credits ConsumedThe total is deducted from your overall balance02 — Details
03 — Optimization (1/2) 03 — Optimization (2/2) 04 — Governance
05 — PricingVoir les détails Voir les détails Voir les détails Contactez-nous 06 — FAQ
Cost Center Official Swiftask Guide
The Swiftask Cost Center is the financial governance system for your AI usage. A credit is the universal unit that measures every AI operation on the platform.
In Summary
Consumption depends on the chosen LLM model, the length of the conversation, and the number of active skills. Savings of up to 40% are possible by applying the best practices described on this page.
They trusted us
How is your credit calculated?
One credit = one unit of AI power. Here is exactly what makes up each interaction with a Swiftask agent.
Prompt
Message & ContextDocuments
OCR & parsingHistory
Previous MessagesInstructions
System PromptSkills
Activated SkillsMultiplier
LLM Model CostCredits ConsumedThe Impact on Your Bill
2.Conversation Length
Each new message includes the entire previous history. Beyond 20 messages, the cost per message skyrockets. Start a new session as soon as the context changes.
Medium-High Impact
3.Active Agent Skills
Each activated skill consumes additional credits with every call. Only activate what the agent strictly needs for its primary mission.
Swiftask: Starting from €18
10 Rules to Never Waste a Credit
Drawn directly from the experience of our clients who have reduced their costs by up to 40%.
Practice #01
New Session
Beyond 20 messages, the history weighs heavily on each request. Close the conversation and start over: your agent doesn't need to remember everything.
Practice #02
LLM Model Suited to the Task
Premium models (Claude, ChatGPT) only for complex tasks. Summaries, FAQs, rewording: default Swiftask model. This single action can reduce costs by 40%.
Practice #03
One Agent = One Single Objective
A generalist agent consumes more and produces less accurate results. Create specialized agents: support, analysis, writing, research.
Practice #04
Minimize Active Skills
Each activated skill = additional cost per call, even if it's not used. Deactivate everything that isn't essential to the primary mission.
Practice #05
Short and Precise Prompts
A vague prompt generates costly back-and-forth exchanges. A clear prompt delivers the right answer on the first try. Remove any repetition and unnecessary context.
10 Rules to Never Waste a Credit
Drawn directly from the experience of our clients who have reduced their costs by up to 40%.
Practice #06
-25% with Annual Billing
If Swiftask is embedded in your organization, switch to annual billing. The savings on the Pro plan can reach up to 25%. It's the simplest discount there is.
Practice #07
Limit Web Scraping
A knowledge base import at depth 3+ consumes massively. Limit it to depth 1 or 2. Upload targeted documents rather than scraping an entire site.
Practice #08
Split Large Files
Uploading a 200-page document all at once: costly and inefficient. Split it into thematic sub-sections and create specialized agents per domain.
Practice #09
Shared Prompt Library
Prompts that work → save them in the Swiftask library. Share them with the entire team to avoid unnecessary iterations.
Practice #10
Mandatory Monthly Review
Analyze the consumption dashboard at least once a month. Identify agents over budget, overused models, and adjust accordingly.
Cost Control & Monitoring
Swiftask Cost Control gives you all the levers to manage spending in real time, without slowing down your teams.
Take control
Daily budget per agent
Set a spending cap. The agent stops gracefully when the limit is reached. Resets every night.
Monthly spending cap per member
Prevent a single user from consuming the entire team budget. The admin sets individual quotas.
Proactive threshold alerts
Receive an in-app notification before reaching your limit. Set a threshold per agent.
Monitoring & Analytics
Personal usage
Each user sees their consumption in real time: by agent, by LLM model, and by day.
360° workspace view
The admin sees who is consuming what, with a complete audit log and automatic alerts.
Reporting & Alerts
Proactive email alerts before reaching your thresholds. Export billing data.
The right plan for your usage
Annual billing is the simplest cost-reduction strategy. Save up to 25% on the Pro plan.
Pro
Parfait pour l'entrepreneur solo
18€
1 Utilisateur / mois Facturé annuellement
Team Starter
Popular
Parfait pour les petites équipes
129€
10 Utilisateurs / mois Facturé annuellement
Team Growth
Pour les équipes en croissance
649€
50 Utilisateurs / mois Facturé annuellement
Entreprise
Obtenez une solution d'entreprise personnalisée
Frequently Asked Questions
Three main factors influence consumption: the chosen LLM model (premium models like Claude or ChatGPT cost more), the length of conversations (the history accumulates with each message), and the number of active skills on your agents, even if they are not used on every call.
Apply three immediate actions: switch to the default Swiftask model for routine tasks (around -40%), start a new session after 20 messages, and disable non-essential skills on each agent.
No, thanks to Cost Control. The admin can set monthly caps per user and daily budgets per agent. Once the limit is reached, the agent stops gracefully without blocking other team members.
No, the credits included in your plan are renewed each month and are not carried over. That’s why it’s important to size your plan accurately and purchase additional credits as needed.
The cost varies depending on the model multiplier. The default Swiftask model is the most cost-efficient. Premium models (Claude Sonnet, GPT-4, Gemini Ultra) have a higher multiplier. Use them only for complex tasks requiring advanced reasoning.
Each user has a dashboard showing real-time consumption: by agent, by LLM model, and by day. Admins have a 360° workspace view with a complete audit log and can export the data.
No. The system sends proactive alerts (in-app and by email) before reaching the configured thresholds. The user is notified in advance and can adjust their usage. Blocking only occurs once the limit is actually reached.
Yes. Workspace administration allows you to define separate budgets and track consumption by usage type. You maintain full visibility over credits used internally versus those consumed by public agents.
Your AI under control Starting today
Access your consumption dashboard and start optimizing in 5 minutes