Skip to content

Trust · security, privacy, money

Clear limits. Clear prices.

How Spiron protects your secrets and data, keeps your workspace separate, and keeps spending predictable — stated plainly, without claims we cannot back up.

Website
https://spiron.ai
API
https://api.spiron.ai
Security contact
security@spiron.ai
Plans
Free $0 · Pro $20 · Max $100
On this page10 sections
  1. 01What you can count on
  2. 02When it asks you
  3. 03Access and secrets
  4. 04Workspace isolation
  5. 05Prices and spend caps
  6. 06Your data
  7. 07AI models
  8. 08What we do not do
  9. 09What we do not claim
  10. 10Contact

Commitments

Six commitments.

They apply on every plan.

Spiron's commitments
CommitmentWhat it means in practice
Work is never lostA restart, an outage, or a long wait does not lose the job or its progress.
Actions are not duplicatedA retry does not send the same email twice or issue the same refund twice.
Every important action is recordedYou can see what was done, when, and what the other system confirmed.
Failures are reported as failuresIf an action did not go through, you are told — it is not counted as done.
It stays inside your limitsIt uses only the tools, data, and budget you gave it.
Your data stays in your workspaceMemory, conversations, and credentials are scoped to your organization.

When it asks you

It works on its own, and asks when it should.

Like a good colleague: it does not ask permission for routine work, and it does not guess on decisions that are yours.

Routine work
Handled without interrupting you.
Important or irreversible actions
Paused for your approval, if you choose to require it.
Unclear situations
Handed to a person with the full context attached, instead of a guess.
Progress
Reported along the way, and quiet when there is nothing to say.

Access and secrets

Your secrets stay secret.

Spiron can use your API keys and tokens to do its job without ever showing them to the AI model.

How access and secrets are handled
AreaHow it works
Secrets are never shown to the AIThe AI only sees a secret’s name. The value is added to the request at the last moment and never appears in prompts, logs, or search.
Sharing is explicitLetting another agent or teammate use a secret is an explicit permission; the secret itself is not copied.
Every use is loggedA secret is only provided to the step that asks for it, and each use is recorded.
Locked secretsYou can lock a secret so an agent cannot overwrite it.
Permissions checked every timeAccess is checked on every request against your workspace’s rules.

Workspace isolation

Your work stays in your workspace.

Isolation is built into every job, not added as a filter afterwards.

Separate by default
Every job, memory entry, and credential belongs to one workspace.
No cross-workspace access
An agent cannot see other organizations’ data, or search beyond what you gave it.
Sub-tasks inherit limits
Work an agent hands off keeps the same permissions — never more.

Prices and spend caps

Published prices, with caps.

Every workspace pays the same published price. Each plan includes AI usage, and work pauses at the allowance.

Plan prices, included usage, and spend caps
PlanPriceIncluded usageExtra usage
Free$0 / month$3 per monthNone: work pauses, no bill
Pro$20 / month$30 per monthNone: work pauses, no bill
Max$100 / month$180 per monthNone: work pauses, no bill
All prices in USD.
One list rate
Included and extra usage are counted at the same list rate, and paid plans include more usage than they cost.
Clear bills
You see your usage and your allowance — broken down by tokens.
Real caps
Work pauses at the allowance, so a paid plan never bills more than its price.

Your data

What is stored, and how you remove it.

We store what your agents need to do the work you asked for, and give you the tools to remove it.

How your data is handled
AreaHow it is handled
SecretsEncrypted at rest and never written into prompts, logs, or conversation history.
ChangesEvery change to a stored credential is saved together with its audit record.
DeletionWhen you erase data, running work is stopped first, so nothing can write it back.
PrivacyPrivacy redaction runs when context is assembled for the AI.
This websiteNo third-party analytics, advertising, or tracking scripts.

AI models

Models do the thinking. You set the rules.

What a model decides
How to approach the next step — not whether the job is done, and not what you allow.
What a model can reach
Only the tools you allowed. Never a raw password or key.

What we do not do

Deliberate limits.

Never

  • Show your secrets or passwords to the AI model.
  • Act outside the permissions and budget you set.
  • Approve its own high-impact changes.
  • Report a job as done without evidence.
  • Bill a Free workspace for overage.
Things Spiron deliberately does not do
We do notWhy
Let an agent approve its own high-impact changesImportant changes need someone other than the one who made them.
Report success without evidenceA job is done only when the result is confirmed.
Put your secrets in front of the AIWhat the model never sees, it cannot leak.
Bill you estimatesEvery charge comes from measured usage.

What we do not claim

What this page does not promise.

We would rather tell you what we do not claim than fill the gap with adjectives.

Questions this page does not answer with a guarantee
QuestionOur answer
Do you guarantee uptime?We do not publish an uptime guarantee today.
Which certifications do you hold?We do not claim any security certification on this page.
Who else uses Spiron?We do not name customers without their agreement.
How long is my data kept?We do not publish fixed retention periods today. You can erase data from your workspace at any time.
How would you tell me about an incident?We do not publish a notification commitment on this page. Report anything you find to the security contact below.

Contact

Security contact and responsible disclosure.

One address, monitored, for vulnerabilities and security questions.

Security

security@spiron.ai

Report vulnerabilities to security@spiron.ai. Include what is affected and enough detail for us to reproduce the issue.

Please do not post secrets, customer data, or a working exploit in public. If you think a credential has leaked, rotate it first, then tell us.