01 / ISOLATE
Workspace boundaries by design
Give every team a bounded environment for agents, tools, knowledge, policies, and compute.
Loading APIToolz
Preparing the application shell.
Loading APIToolz
Preparing the application shell.
APIToolz is the control plane for running isolated AI workspaces, specialized agents, permission-aware knowledge, governed tools, and approval-driven execution.
Built on the infrastructure your team already trusts
Stop stitching together agents, queues, permissions, and logs. APIToolz gives operators a coherent path from user instruction to auditable result.
Give every team a bounded environment for agents, tools, knowledge, policies, and compute.
Route every external action through permissions, approvals, rate limits, and output guards.
See task state, runtime health, queue pressure, reasoning traces, and audit evidence in one place.
KNOWLEDGE, WITH A CLEAR PATH TO ACTION
Turn policies, handbooks and product documents into permission-aware answers. Keep people in control of what gets published—and what happens next.
FOLLOW THE KNOWLEDGE LIFECYCLE
Illustrative walkthrough · no live dataRetrieve within the reader’s permissions, check the evidence, and make the source easy to inspect.
Apply workspace, knowledge base and selected-document restrictions.
Workspace permissionsCombine semantic and keyword retrieval from current approved sources.
Qdrant + metadata filtersCheck support and qualifications. Withhold unsupported answers.
Configured AI providerRead citations and inspect the relevant page, sheet rows or excerpt.
Portal / configured TelegramNo unverified answer text is streamed. Publication and access are checked again before delivery.
BEYOND FILE UPLOADS
Bring PDFs, Office files and text. Inspect extracted passages, candidate tables and English/Burmese OCR before approval.
OCR languages need installed parser packsKeep document history, effective dates and independent publication review. Only eligible current versions can answer.
Write source-backed FAQs, categorize content and offer reviewed suggested questions. FAQs follow the same publication checks.
Save encrypted portal conversations. Confirm contextual follow-up questions, then search current sources again—not past AI answers.
Opt-in · history encryption and retention requiredTrack indexing, document owners, review dates and answer feedback. Assign issues and inspect privacy-controlled diagnostics.
Schedule imports from allowlisted S3-compatible sources with read-only credentials. New and changed content becomes a draft.
Opt-in · source credentials and schedule requiredONE KNOWLEDGE BASE. CONTROLLED ACCESS.
Use the authenticated portal, or explicitly connect a public KB to an allowlisted Telegram chat. Public Q&A stays separate from private applicant information and secure Mini App collection.
Plan your knowledge workspaceThe orchestrator decomposes work, selects the right agent, and routes execution through your security and approval model—before a tool ever touches production.
Turn a human instruction into explicit, reviewable work.
Assign specialized agents and route tasks through resilient queues.
Enforce permissions, policy, approval, and secrets at the tool boundary.
Capture logs, decisions, runtime evidence, and every state transition.
Launch with the agents, tools, guardrails, and runtime shape your domain needs—then evolve it without rebuilding the platform.
GitHub · Docker · PostgreSQL
Use this playbookMikroTik · FreeRADIUS · SMS
Use this playbookGrafana · Kubernetes · Terraform
Use this playbookClean architecture. Portable infrastructure. No mystery control plane between your agents and your business.
Review the system design