● Available now AI-Native Clients & Internal teams GDPR by design Embeddable

Every business request.
AI manages it. Autonomously.

CREIMSCustomer Request & Enhancement Intelligent Management System — is the AI-Native platform that governs any type of request: bug reports, suggestions, support, change requests, and enhancement proposals — from external clients and internal teams alike. The AI operates autonomously end-to-end, escalating to humans only where judgement or policy require it. Standalone deployment. No full AGI-ONE platform needed.

Bug · Suggestion · Support · Change · Enhancement  |  C-One widget · Grafana alerts · Jira · ServiceNow · Plane.so

Any request type & source
AI-Native autonomous triage
Full lifecycle governance

Every unmanaged request is a cost. And they multiply.

Bug reports, suggestions, support cases, change requests, enhancement proposals — every type of request handled manually means lost time, lost context, and lost value. For your clients and your internal teams.

Manual and slow classification

Every request — bug, change, or enhancement — is read, classified, and prioritized by hand. Most of it has a known answer or a clear owner already.

Fragmented channels

Requests arrive via email, chat, forms, and Slack. No single system governs them all — clients and internal teams use different tools with no shared view.

Zero institutional memory

Fixes, decisions, and enhancement rationale live only in someone's inbox. Every new request starts from zero — the same problems recur, the same answers are re-discovered.

No traceability or ownership

Enhancement requests disappear in backlogs. Support cases have no owner. Change requests don't have a status. No audit trail, no accountability, no SLA.

All of this is solved by a single autonomous system — governed at every step, with no black boxes. It reduces implementation costs and resolution times by over 50%, eliminates dependency on individual knowledge, and builds its own understanding of your software and systems autonomously — so the next request is handled better than the last.

An AI that governs. Not just classifies.

CREIMS is the AI-Native engine that manages the full lifecycle of any business request — from intake and triage to resolution and Team Work handoff. C-One is the embeddable widget your users and teams interact with. Together they replace the fragmented, manual process with a single governed flow.

Frontend · C-One Widget

The intake channel
for every request type

C-One is the embeddable widget that sits on any web portal — client-facing or internal. It guides users through submitting any type of request: bug report, suggestion, support case, change request, or enhancement proposal. Automatically collects full page context, generates an AI summary, and feeds CREIMS.

  • 5 request types: Bug · Suggestion · Support · Change · Enhancement
  • JS/CSS on-demand loading — zero performance impact on host portal
  • AI conversation summary + user final review before submission
  • Real-time staff notification via email, Telegram, and WhatsApp
  • Voice input via browser Speech Recognition
💬 C-One Support

Hello! How can we help you?
Choose a request type to start

🐞
Report Bug
Report a problem or error encountered
Request New Agent/Service
Request a relevant new feature
💡
Suggest KB Content
Propose new knowledge base content
📋
Support
Any other needs not included above
Backend · CREIMS — AI Support Engine

The AI-Native engine
governing every request

CREIMS is the autonomous governance engine for the full request lifecycle. It classifies, triages, proposes solutions, manages escalations, and drives handoff to execution — for any request type, from any source (C-One widget or Grafana alert). The AI operates end-to-end. Humans intervene only where policy or judgement require it.

Multi-tenant KB-API

PDF/DOCX/MD upload, semantic chunking, vector embedding, hybrid search per client.

Auto-Triage Service

AI analysis, fix proposal or escalation for every ticket — with semantic matching of the best worker based on ticket content and worker skills and abilities.

Grafana Webhook + Alerting

Ingestion of infrastructure alerts from Grafana, automatic conversion to trackable support requests.

Continuous feedback loop

Each resolution enriches the error knowledge base. The system learns from every outcome and improves autonomously.

From any request to governed resolution.

An end-to-end dev cycle where AI handles classification, triage, and routing autonomously. Human intervention is reserved for cases where judgement, policy, or explicit validation add real value. The same flow applies to client requests and internal team requests alike.

1
Request enters CREIMS — from two sources
Source A — C-One Widget: the embedded widget guides the user through selecting a category and describing the issue. Automatically collects URL, page, section, timestamp, and browser metadata.
Source B — Grafana / Alertmanager: infrastructure alerts (LogQL on Loki, PromQL on Prometheus) fire automatically and are ingested via webhook. CREIMS deduplicates by Grafana fingerprint before creating a Request.
Human: widget Auto: Grafana alert
2
C-One generates summary and creates Request
For widget-submitted requests: C-One generates a professional AI summary of the conversation. The user reviews and confirms before submitting. For Grafana alerts: the webhook payload is normalized automatically. In both cases the Request is persisted in support_db with status sent and staff is notified via email, Telegram, and WhatsApp.
Auto: AI summary Human: submit review
3
CREIMS analyzes and classifies automatically
The CREIMS Support Agent polls Requests in sent status, brings them to in_analysis, queries KB, error history, and known patterns. The Detector Agent applies rule-based dedup and pattern matching in parallel. Decides the path autonomously: proposed fix, triaged, duplicate, escalated.
Fully autonomous
4
Handoff to Team Work — AI executes, humans review
Ready Requests are transferred to Team Work, which semantically matches the best worker based on ticket content and worker skills. For software requests, the Development Executor invokes Claude Code or Codex CLI autonomously: it clones the repo, implements the fix on a dedicated branch, and opens a GitHub PR. All work items are tracked through progress checkpoints toward UAT.
AI: autonomous execution Human: go-ahead & PR review
5
Human supervision — two precise checkpoints
Everything between intake and closure runs autonomously. Human intervention is confined to two explicit moments.

Checkpoint 1 — Work start approval: before the assigned AI worker begins execution, the Developer Control Room surfaces every detail — request context, proposed approach, assigned worker, and execution parameters. The operator reviews and gives the explicit go-ahead. Nothing starts without this confirmation.

Checkpoint 2 — UAT validation: once the work is complete, the operator validates the result in the UAT dashboard, reviews the opened PR and the changes, then approves or rejects with traceable rework notes.
AI: fully autonomous between checkpoints Human: go-ahead & UAT approval
6
Release Management & post-deploy validation
Tasks approved in UAT flow into the Release Management dashboard, where operators can review everything ready for production in one view — grouped, prioritised, and traceable back to the original requests. Each item can be scheduled for a production release: the system executes the merge and updates the production repository at the defined time.

After deployment, an automated post-deploy agent runs verification checks directly in production, signals any anomaly with full context, and links it back to the release record — closing the loop from original request to verified delivery.
AI: scheduled deploy & post-deploy testing Human: release scheduling Coming soon
sent
in_analysis
triaged
fix_proposed
escalated
assigned
in_delivery
waiting_uat
resolved
release_scheduled
deployed
closed

One system. Every request type. AI-Native.

CREIMS is not an AI layer on top of a traditional ticketing system. It is built AI-Native from the ground up — the AI is the operating engine, not a feature. These are the operational capabilities available today.

5
Docker backend services
2
Automatic detection sources
5
Request types
10+
Tracked lifecycle states
3
Staff notification channels

C-One Embeddable Widget

Two JS/CSS files for any web portal — client-facing or internal. Covers all request types: bug, suggestion, support, change, enhancement. Zero dependencies, opens on-demand.

AI-Native Autonomous Triage

The AI classifies and routes every request type autonomously — bug, enhancement, or change — consulting KB, error history, and request patterns. It proposes solutions, detects duplicates, and semantically matches the best worker based on ticket content and worker skills and abilities. Escalates only where human judgement is needed.

Multi-tenant KB

Upload corporate documents (PDF, DOCX, MD). Semantic chunking, vector embedding, hybrid search. Each client has their own isolated KB.

Error Pattern Learning

A graph-based error knowledge base records every resolution and known fix. The system learns continuously and improves future triage without manual configuration.

Multi-channel Notifications

Every Request notifies staff via email, Telegram, and WhatsApp through the Communication Gateway v2. Configurable by channel, recipient, and request type.

External Ticketing Integration

Unified microservice bridging CREIMS requests to Jira, ServiceNow, ClickUp, and Plane.so. Credentials managed via HQ UI, not in environment variables.

Detector Agent (v2)

A dedicated BE-only agent running rule-based detection and duplicate identification in parallel with the Support Agent. Reduces false positives, deduplicates ticket floods, and flags anomalous request patterns before human review.

Grafana & Alertmanager Integration

CREIMS automatically ingests infrastructure alerts via Grafana Unified Alerting or Alertmanager webhook. Supports LogQL (Loki) and PromQL (Prometheus) rules. Native deduplication by Grafana fingerprint prevents ticket flooding from persistent alerts.

3-step integration.

From widget to ticket resolution in minutes.

Widget on any site

Portals, intranets, web dashboards: just 2 HTML tags to add the AI support channel to any page.

Docker Compose ready

Full CREIMS stack (5 containers) deployable with a single command. On-prem, cloud, or hybrid. Dedicated port, reverse proxy included.

No vendor lock-in

Data in standard PostgreSQL. Documented FastAPI. No proprietary formats. Export, migrate, or integrate with any system.

Scales to AGI-ONE

When you're ready, CREIMS connects natively to the full platform. No migration needed: it is already the correct subset.

Embed widget on your portal — 2 lines
<!-- 1. Load the widget (on-demand, zero blocking) -->
<script src="https://your-domain/c-one.js" data-api="https://your-creims-host:8190" ></script>

<!-- 2. Widget styles (do not affect the host site) -->
<link rel="stylesheet" href="https://your-domain/css/user-support-assistant.css" />

<!-- The widget appears as a floating button. Nothing else to configure. -->

Team Work: the execution engine. ● Active

Once CREIMS classifies and validates a Request, Team Work takes over the full execution lifecycle — assigning the right worker (human or AI agent), tracking progress, running UAT, and closing the loop back to CREIMS. For software tasks, AI workers can execute code changes autonomously and open pull requests ready for human review.

Semantic Worker Matching

AI selects the most competent worker for each task using vector similarity on worker profiles — not rule-based routing. Human and agentic workers in the same pool.

Development Executor

For software tasks, AI agents (Claude Code, Codex) autonomously clone the repo, implement the fix on a dedicated branch, and open a GitHub PR — ready for human review. The merge is always human.

UAT & Governed Closure

Every task passes through waiting_uat: a designated operator validates the result, approves or rejects. Approved tasks close the CREIMS Request. Rejected ones return to the worker with traceable rework context.

Resume Checkpoint

Every task records its last completed step, artifacts touched, and next action. Any interruption — human or agent — resumes from exactly where it stopped, with full context for the successor.

Team Work is operational.  Deployed on Docker with full API, admin console, and CREIMS integration active. The Development Executor supports Claude Code and Codex CLI as autonomous agentic workers. Contact us to see it in action.

Frequently asked questions about CREIMS

CREIMS — Customer Request & Enhancement Intelligent Management System — manages any type of business request: bug reports, suggestions, support cases, change requests, and enhancement proposals. It serves both external clients (via the C-One widget on customer portals) and internal teams (via internal portals or direct integrations). All request types flow through the same AI-Native governance engine, with lifecycle states, KB lookup, and Team Work handoff.

No. CREIMS and the C-One Widget are designed to work independently. The backend stack deploys with a single Docker Compose (5 containers). The C-One widget is a JS/CSS inclusion on any web page. No other components of the AGI-ONE platform are required to start.

Yes. The system is on-prem by design: Request data remains in your PostgreSQL database, under your control. No data flows to unauthorized cloud services. Provider credentials are managed via UI and never in plain text on config files. The system is designed for GDPR-compliant environments from the architecture up.

Both. CREIMS has two request sources: (1) C-One Widget — users submit reports from any web portal; the widget guides them through category selection and collects full page context. (2) Grafana / Alertmanager — infrastructure alerts (container down, error rate spike, latency threshold) fire automatically via webhook and are converted into CREIMS Requests with zero manual input. Both paths feed the same Auto-Triage engine for consistent governance.

CREIMS runs two agents in parallel. The Support Agent is conversational: it interacts with the user through the C-One widget, collects context, and generates the AI summary. The Detector Agent (v2) is backend-only: it runs rule-based analysis on incoming requests, detects duplicates, flags anomalous patterns, and pre-classifies tickets before the Support Agent processes them. Together they reduce triage latency and false positives significantly.

Currently, CREIMS uses Google Gemini 2.0 Flash via Google AI API and the Google ADK (Agent Development Kit) runtime. The system is designed to be model-agnostic: the model choice is configurable and can be adapted to local models or other cloud providers without rewriting the triage logic.

Yes. The 5 default request types (Bug, Suggestion, Support, Change, Enhancement) are the starting point, mapped to the CREIMS taxonomy system. Categories, labels, initial messages, and collection flows are all configurable. For advanced customizations — new types, additional fields, vertical-specific workflows — we support the client in tuning the widget and CREIMS mapping.

Yes. CREIMS includes a dedicated Ticketing Integration microservice with a unified API supporting Jira, ServiceNow, ClickUp, and Plane.so. Phase 1 is read-only (sync and status mapping); Phase 2 enables creating and updating remote tickets directly from CREIMS. Credentials and provider configuration are managed through the HQ UI — never in environment files.

The system moves the Request to an escalated state. The CREIMS operator receives a notification, sees the full context (text, user, page, category, AI summary, history), and can provide the missing information to restart the triage. No Request is lost: the escalated status is visible and trackable in the dashboard.

Ready to govern every request with AI?

One system. Any request type. Clients and internal teams. AI operates. You supervise.