Schema isolation
Structured Data
Users, metadata, and configuration in a dedicated schema. No shared tables.
Security
Decide what an agentmay see and do.
- Agent 365 evaluates inside Microsoft.
Identity, scope, and credentials resolve on the call — before the write, in your tenant.
GitHub
- Read repository Allowed
- Open pull request Approved
- Force-push main Blocked
Jira
- Read issue Allowed
- Create issue Approved
- Delete project Blocked
Linear
- List issues Allowed
- Update status Approved
- Archive team Blocked
Slack
- Read channel Allowed
- Post message Approved
- Invite guest Blocked
tool
Resolve.ai
team
Platform eng
team
On-call
tool
Analyst
user
Maya Chen
tool
SRE bot
team
Security
user
Jordan Lee
tool
Docs agent
Actions gated per integration
Security Model
Discovery. Policy. Record.
The gate in front of the write — architectural, not a vendor-owned record.
Estate discovery
See the agents and clients already in the organisation before you gate them.
Policy at execution
Identity, scope, and short-lived credentials resolve on the tool call — before the write.
Decision record
Each agent has a defined read scope, write scope, and approval gates. Every decision is recorded.
Infrastructure
In practice.
Every call authenticated. Every store isolated.
Credential path
A service can only mint the next hop.
MCP Gateway
tenant identity + role
Knowledge Service
scope: read:index
Indexing Service
scope: read:source
Integration Service
github · create_pr
Token at hop 3
41s left- sub
- svc/indexing-service
- aud
- svc/integration-service
- scope
- call:github
- tenant
- t_northwind
cannot mint write:*
Not in the parent scope. A service cannot widen its own grant.
Tenant boundary
Separate storage by design
Index isolation
Semantic Index
A separate index per tenant. Another org's query cannot reach your content.
Space isolation
Connection Map
Your entity graph in a dedicated space — structural separation, not a query-time filter.
Namespace isolation
Cache & Sessions
Namespaced per tenant. Encrypted at rest. TTLs enforced.
Permissions
Per-agent IAM.
Defined read, write, and approval gates. Enforced at runtime.
AgentRead scopeWrite allowedApproval gated
Incident responderIncidents + linked PRs/tickets/servicesCreate incident comments, update status
Acknowledge + escalate
Code quality agentPRs, commits, issues across reposCreate PR comments, labels
Close issues, approve PRs
Review agentPR diff + code context + similar patternsPost review comments
Approve / request changes
Security agentRepos, dependencies, CVE dataCreate security issues
Merge or close security PRs
IAM agentIAM policies, permissions, access logsFlag policy violations
Modify any permission grants
Onboarding agentAll public org docs + onboarding materialsUpdate onboarding docs
Write to protected spaces
Approval gates require human confirmation. Every action, approved or not, is audited.
Encryption everywhere
AES-256 encryption at rest across all data stores
TLS 1.3 for all data in transit, inter-service and client
Credentials stored as encrypted secrets, never in plaintext
All indexed data encrypted per tenant in isolation
Auth tokens are short-lived, signed, and tenant-scoped
Self-hosted option (roadmap)
Not yet offered. The platform already ships as Helm charts, and customer-cloud delivery on your own GCP, AWS, or on-prem is planned so data stays inside your network. Portable permissions, knowledge, and sessions staying in that deployment is sovereignty.
Helm chart deployment on Kubernetes
All services: Knowledge, Indexing, Agent Runtime, Gateway
Bring your own cloud credentials and API keys
Annual software license with no per-query cost to Telara
Same security architecture as cloud, no shared infrastructure
Observability
Audit & observability
Every action logged. Tamper-evident. Retained to your spec.
Full audit trail for all agent actions and queries
Immutable event log: actor, action, resource, timestamp
Exportable to your SIEM (Datadog, Splunk, CloudWatch)
Approval gate decisions logged with approver identity
Permission violations trigger automatic alerts
Standards
Compliance status
SOC 2 Type II
Controls implemented; audit planned
HIPAA
Planned
GDPR
Processor-aligned
Self-hosted
On the roadmap
Security review and control details available on request.