NewEvery agent tool call now routes through the governed MCP gateway — policy check, audit log, then dispatch.Read the changelogarrow_forward
XOOZLYAI Orchestrator
routeHow it workssmart_toyAgent fleetshieldGovernance & trustsellPricingmenu_bookDocsarticleBloghistoryChangelog
Log inStart free
grid_viewProductsrouteHow it workssmart_toyAgent fleetshieldGovernance & trustsellPricingmenu_bookDocsarticleBloghistoryChangelogHelp centreLog in
Start free
  1. XOOZLY
  2. chevron_rightDocumentation
Loading the knowledge base…
XOOZLYAI Orchestrator

Ready to orchestrate your AI fleet with total confidence?

Start a project, set a cost cap, and watch the loop run with every tool call policy-checked and logged before it leaves the platform.

Start freearrow_forwardSee what shipped
  • check_circleNo credit card required
  • check_circleYour keys, your model spend
  • check_circleEvery tool call audited

Platform

  • All productsarrow_forward
  • Orchestration looparrow_forward
  • Agent fleetarrow_forward
  • Governed control planearrow_forward
  • Pricingarrow_forward

Trust & legal

  • Trust centrearrow_forward
  • Security & compliancearrow_forward
  • AI use policyarrow_forward
  • Report a vulnerabilityarrow_forward
  • Data subject requestsarrow_forward

Resources

  • Documentation & guidesarrow_forward
  • Engineering & research blogarrow_forward
  • Changelogarrow_forward
  • Uptime & reliabilityarrow_forward

Browse by domain

  • Strategy & Researcharrow_forward
  • Developmentarrow_forward
  • QAarrow_forward
  • Governance & Orchestrationarrow_forward
  • All 10 domainsarrow_forward
XOOZLY
© 2026 XOOZLY · 61 agents · 10 domains · 8 SDLC stationsAI use policyarrow_forwardLog inarrow_forward99.9% uptime SLAarrow_forward
  1. XOOZLY
  2. chevron_rightDocumentation

Public knowledge base · no account needed

Documentation & technical guides

Guides to running a governed agent fleet on XOOZLY: cost caps checked before a run starts, MCP connectors behind one audited gateway, versioned rule files, tenant isolation and execution traces.

search⌘K
Popular topics

16 of 16 guides

support_agentSigned in? Search the in-product Help Centrearrow_forward
play_circle

Core concepts & getting started

4 guides

CORE
  • Setting a per-project cost capA project's cap is a hard limit in integer cents, checked before a run starts rather than reconciled afterwards. A breach pauses the project; only a Founder/Admin resumes it.articleGuide · opens Set a cost cap in the consolearrow_forward
  • Connecting your first MCP serverRegister a server, store its credential encrypted, pass the safety sweep, then grant an agent the exact tool actions it may call on that connection.articleGuide · opens Connect an MCP server in the consolearrow_forward
  • Understanding run states and lifecycle transitionsEvery run moves queued → planning → executing → completed, failed or paused. One run is one pass through stations A–H; the loop back to A is the next run, not a restart.articleGuide · opens Run log in the consolearrow_forward
  • How rule files govern an agent, and how a version is approvedAn agent's behaviour is governed by its rule file. Versions are append-only: a change writes a new row rather than editing one, so every previous version stays rollback-ready.articleGuide · opens Agent fleet in the consolearrow_forward
hub

MCP protocols & integrations

4 guides

INTEG
  • The MCP gateway: policy check, audit log, then dispatchEvery agent tool call takes one path: policy check, then an audit record written whether it passed or failed, then connector dispatch. There is no bypass, and no implicit allow.articleGuide · opens MCP gateway policy editor in the consolearrow_forward
  • Importing and exporting MCP connector configurationsMove a project's connector configuration between environments as a reviewable bundle. Credentials never travel with it — each environment holds its own.articleGuide · opens Connector config export & import in the consolearrow_forward
  • Webhook signature verification and retry handlingRegister an endpoint, verify the signature on what you receive, and follow each delivery attempt. Outbound delivery retries three times with exponential backoff before it is marked exhausted.articleGuide · opens Webhooks in the consolearrow_forward
  • OAuth callback flows for third-party developer toolsInspect a failed authorization round trip: the redirect URI the provider actually received, the state and PKCE values, and the error it returned.articleGuide · opens OAuth callback debugger in the consolearrow_forward
security

Governance, compliance & security

4 guides

GOV
  • How one project's data stays isolated from every otherTenant-owned rows carry their own project or org id, every query filters on it explicitly, and Postgres row-level security sits underneath as defence in depth — never as a replacement for the filter.articleGuide · opens Trust centre in the consolearrow_forward
  • Data subject requests: intake, hold and what erasure removesIntake, legal hold and fulfilment for a GDPR/CCPA request, plus exactly what an erasure removes — and what the append-only audit log keeps, redacted, because it is never deleted.articleGuide · opens Data subject requests in the consolearrow_forward
  • Managing organization custom roles and permission gatesFour default org roles cover most teams; Enterprise adds custom roles built from permission groups. Every check runs server-side against the caller's real membership, never a client-supplied role.articleGuide · opens Custom roles in the consolearrow_forward
  • Setting up TOTP MFA, and where security keys standEnrol a TOTP authenticator and keep its backup codes. Security-key enrolment is also available, with its signature verification still to come — so TOTP remains the factor to rely on today.articleGuide · opens Enrol a second factor in the consolearrow_forward
monitoring

Agent fleets & observability

4 guides

OBS
  • Reviewing Rule Optimizer automated proposalsThe optimizer proposes; a person approves. Approval is what creates the new rule-file version — there is no path where a proposed change takes effect on its own.articleGuide · opens Approval queue in the consolearrow_forward
  • Interpreting execution traces in the Trace ViewerOpen a completed cycle to read its trace: every station's output, timestamped, with the tool calls a station made and the cost each one booked.articleGuide · opens Cycle history in the consolearrow_forward
  • Failed step diagnostics and human-in-the-loop corrective actionsA failed station does not retry itself in silence — it surfaces for evaluation with the state it actually produced, so a person decides whether to correct, rerun or stop.articleGuide · opens Orchestration dashboard in the consolearrow_forward
  • Long-running asynchronous agent execution trackingAgent runs and bulk operations are queued work, not blocking requests. Follow a long job's progress, see where it is held up, and find the ones that failed.articleGuide · opens Global jobs dashboard in the consolearrow_forward

How the governance layer works

Every tool call leaves a record

An agent cannot reach a connector directly. Each call is checked against that connection’s own policy — a grant naming the exact actions that agent may take — then written to the append-only audit log whether it passed or failed, and only then dispatched. A write is not recorded as done until it has been read back.

3-step gateway
Policy check, audit record, then dispatch — in that order, with no bypass path
8 SDLC stations
Stations A–H, routed by the Governance & Orchestration hub
Read the engineering write-upsarrow_forward
audit_logs / mcp_tool_call
{  "action": "MCP_TOOL_CALL_ALLOWED",  "actor_type": "system",  "resource_type": "mcp_server_connections",  "resource_id": "0199c4a1-7f3e-7c21-9b44-2d6f8ae51c03",  "after_state": {    "agent_id": "0199c4a1-7f3e-7c21-9b44-2d6f8ae51c04",    "action": "create_pull_request",    "correlation": { "runId": "0199c4a1-7f3e-7c21-9b44-2d6f8ae51c05" }  },  "severity": "info"}

One row per call, for every connector type — including the ones with no adapter implemented yet, which return an honest failure rather than a fabricated success.

Deployment & enterprise

Need deployment guidance or an enterprise agreement?

Our engineering team walks through connector scoping, gateway policy design and cost-cap strategy for your fleet. The published uptime commitment and security posture live in the Trust Centre.

Start freearrow_forward