Run Spaceplace where your security rules require: in our EU cloud or on your own servers.
Multi-tenant SaaS: Spaceplace Cloud, hosted on AWS on servers inside the European Union.
Standard
Single-tenant SaaS: Your own Spaceplace instance in the EU, managed by our team.
On request
Bring your own cloud: A dedicated deployment in your own AWS, Azure or Google Cloud account.
On requestOn-premise
On-premise: On your own servers, with your own models and inference.
On request
Harness
Every request passes the same eight checks.
People and agents reach your systems and the models only through the Agentic Harness: one set of rules, checked on every call, whoever makes it.
PeopleEmployees in their workspaces
AgentsAgents and workflows
Agentic Harness
01Identity
Sign-in through your identity provider, SAML or OIDC, with roles from your directory.
02Roles and attributes
Access by role and by attribute, down to the record: a sales agent sees only their division.
03Agent permissions
An agent works with its role's or its person's rights, never more. It can't grant itself access.
04Action classes
Every tool has a reversibility class. Irreversible actions wait for a person to approve.
05Untrusted input
Messages, files and tool results are checked for hidden instructions before an agent reads them.
06Tools and MCP
Only the tools and MCP servers you approve, at pinned versions, with every call logged.
07Model data boundary
You decide which data may reach which model: ours, the providers' or your own inference.
08Audit and export
Every decision is on record, and security events can go to your SIEM.
Your systemsCRM · ERP · telephony · email
ModelsProviders or your own inference
Agent rights
Decide what an agent may do on its own.
Every tool an agent can use has a reversibility class. You set what runs on its own, what waits for a person and who that person is, and what the agent may not do at all.
Support agentRights · set by your admin
ToolsRightsRecord
#ToolReversibilityRuleApprover
1Read the orderRead onlyAuto-approved—
2Read the customer's messagesRead onlyAuto-approved—
3Draft a replyReversibleAuto-approved—
4Change the order statusHard to reverseApproval requiredSupport lead
5Issue a refundIrreversibleApproval requiredHead of supportSupport leadHead of supportFinance director
6Export the customer listIrreversibleDenied—
Support agentRequest · 14:02New requestChecking the ruleWaiting for approvalApproved
Issue a refund · $240Order #48213 · damaged on delivery · tool 5
Rule checkIrreversible, so approval is required
Decision InboxSent to the head of support
ApprovalApproved by Iryna K. · 14:08
RecordWhat the agent saw, the rule and the approver, kept with the decision
1Your admin sets a rule and an approver for every tool2The agent wants to issue a refund3A refund can't be undone, so the rule asks a person to approve4The request goes to the head of support's Decision Inbox5Approved, so the agent issues the refund6Everything is kept with the decision
Monitoring
Real-time monitoring.
See what every agent does while it works: each step, each decision, what it was missing, and what it cost in time and tokens.
01 · SetupYour sign-in and roles02 · ConfigureConnections you approve03 · TestTested on your cases04 · ReleaseEvery decision on record
Build
Take full control of your AI deployment.
01 · SetupConnecting your systems02 · ConfigureSetting up agents and workflows03 · TestChecking answers on real cases04 · ReleaseLive, with your process owner
Forward-deployed engineers
We build it with you
Our engineers connect your systems and build the first agents and workflows with your process owners.
Lead follow-upWorkflow · by your teamSupport agentAgent · by your teamWeekly sales reportReport · by your teamHead of salesRole workspace · by your teamReleaseLive
Your team
Build it yourselves
Your team gets the whole platform, with support and recommendations from our engineers. Build agents, workflows and role workspaces, connect your systems and release when you're ready.
Spaceplace Cloud runs on AWS, on servers inside the European Union. With a single-tenant instance or on-premise, it runs where you choose.
Do you train AI models on our data?No+
No. Your data is never used to train or improve models, ours or the providers'. Model providers process requests under terms that exclude training.
Who at Spaceplace can see our data?Named engineers+
Only the engineers who support your deployment, and only to set it up or fix a problem. We name them in the contract.
Can an agent act without approval?Your rules+
Only where your rules allow it. Each tool has a reversibility class, and anything you mark for approval waits for the right person.
Can we use our own models?Yes+
Connect your own inference as another provider. Evaluation Lab compares its quality and cost with the others on your own cases.
On-premise, does anything leave our servers?Nothing by default+
No telemetry goes to us by default. You can turn it on if you want our engineers to monitor the deployment.
Do you work with Microsoft Entra ID?Single sign-on+
Your people sign in with Entra ID, Okta, Google or any SAML or OIDC provider. Syncing groups and roles from your directory is in development.
Do you hold SOC 2 or ISO 27001?Not yet+
Not yet. We answer your security questionnaire in writing, share our security overview under NDA, and an engineer walks your team through the controls.
What happens to our data if we leave?Exported, then deleted+
We export what you need, then delete your data from every store and confirm it in writing.
What Spaceplace can do for your business
We answer it in writing, and an engineer walks your team through the answers.
We use cookies to see how people use this site and to measure our ads. They stay off until you allow them, and you can change your choice any time in Cookie settings. Privacy policy