Enterprise agent platform
Design for a managed workspace coordinating role-based agents, tools, data access and human-review paths; implementation is engagement scope.
A reusable platform designDigital Futures · Enterprise AI
Standard platform,
bespoke outcomes.
dClaw IAS is designed as a governed enterprise agent platform: a standard foundation for role-based agents and configurable workflows, with tool and data boundaries, human-review paths and evidence requirements scoped for each engagement.
The platform design is paired with bespoke configuration and implementation. Each pilot scope must define and independently verify its controls before customer-facing or customer-data use.
Build status: The 3-tenant, 9-user MVP is in validation. It is not available for customer-facing or customer-data use until the relevant controls and the pilot release gate have been independently verified.
CONCEPTUAL PLATFORM DESIGN
ConfigureAgent · knowledge · workflow design
GovernPolicy · permission · review requirements
EvidenceSource · action · approval requirements
One platform designEngagement-specific configuration
Governed workflow designHuman-review requirementsModel optionsEvidence requirements
01 / Platform
A platform design with controls around the work.
The target design connects specialist agents to scoped tools and data, with policy boundaries, human-review paths and evidence requirements. Components and controls must be implemented and independently verified for an engagement; this diagram is not the current MVP build.
01 · PEOPLEEnterprise usersBusiness and operations teams
02 · ACCESSChat · web · dashboardNatural-language requests
03 · COORDINATIONdClaw IAS orchestrationRoutes work, tools and review
04 · SPECIALISTSRole-based agents
DomainData & queryValidationReport & narrativeAudit
05 · CONNECTED CONTEXTApproved tools & dataDatabases · documents · apps · dashboards
06 · BOUNDARIESPolicies & rulesPermissions · data scope · approval paths
07 · HUMAN REVIEWReview & approvalFor sensitive outputs and high-risk actions
08 · DELIVERYApproved outputReports · briefings · workflow actions
AUDIT EVIDENCE ACROSS THE FLOWPrompts·Sources·Tool calls·Outputs·Approvals·Versions
Technology directionOpen, composable foundations
Go core · Python / FastAPI AI tier · PostgreSQL with Apache AGE and pgvector · Redis · LiteLLM · Caddy · OpenBao · OpenTelemetry / Grafana · Docker Hardened Images
02 / What we deliver
A platform, configured around your operating model.
The commercial offer pairs a standard platform design with customer-specific configuration and implementation support. The capabilities below describe engagement scope, not functions available in the current MVP.
Configurable agent services
Configure business, technical, compliance, reporting and operational agent designs around customer-defined roles and environment.
Customer-specific configurationGoverned workflow automation
Specify approval gates, policy routing, evidence requirements and exception handling for each proposed workflow.
Controls defined for reviewSecure integration layer
Define controlled access patterns for databases, files, applications, dashboards and enterprise systems.
Integration boundaries in scopeHuman-in-the-loop operations
Design reviewer workflows for sensitive actions and outputs; implement and verify an authenticated reviewer path before pilot release.
Human accountability by designObservability & auditability
Define required prompt, action, tool, output, source and version records; build and verify an audit sink before relying on evidence.
Evidence requirements in scopeProfessional services
Professional services can cover solution design, data readiness, workflow configuration, training and change enablement.
A path from design to implementation03 / Capabilities
Target capabilities for governed work.
These are target capabilities from the product brief. Each must be scoped, implemented and independently verified for an engagement before customer-facing or customer-data use.
Natural-language interface
Design for plain-language requests in Vietnamese, English or other approved business languages.
Multi-agent orchestration
A configured target can coordinate retrieval, analysis, validation, reporting, compliance review and follow-up tasks.
Retrieval-augmented knowledge
Scope approved documents, policies, templates, metric dictionaries and knowledge bases as retrieval sources.
Governed data query
Specify data boundaries and review controls for natural-language queries, SQL, charts and dashboards.
Report & narrative generation
Configure report generation for executive summaries, regulatory-style reports, management commentary and briefings, with validation before use.
Validation harnesses
Design validation against benchmarks, prior periods, business rules, source data and reviewer decisions.
Workflow automation
Specify recurring-task, briefing, approval, evidence and exception-routing workflows for implementation.
Audit evidence capture
Define required traceability for requests, sources, outputs and approvals; the full audit sink remains an implementation and release prerequisite.
04 / Design principles
Built around accountable use.
Governed by default
The target design specifies permissions, policies, data boundaries and approval paths for each engagement.
Designed for enterprise needs
Role-based access, audit logging, secure integrations and supportability are design requirements to scope and verify.
Composable
The design aims to reuse a shared dClaw IAS foundation across departments and use cases.
Human accountable
The design requires review before sensitive outputs or high-risk actions are released; an authenticated reviewer path is a pilot prerequisite.
Model flexible
Customer-approved models and private-cloud, on-premise or hybrid deployment are engagement options to evaluate and validate against governance needs.
05 / Where to start
Use cases with a clear operating context.
The examples below are candidate workflows from the product brief. Scope data and review boundaries in design before implementation or pilot use.
01Example workflows
Financial Services
& Banking
- Internal reporting and BI automation
- Management business intelligence reporting
- Regulatory report preparation and evidence packs
- Risk, collections, NPL, DPD and portfolio monitoring
- Policy, compliance and audit assistance
- Customer operations knowledge assistant
02Next increment
Cybersecurity
& IT Operations
- AI-assisted SOC triage and alert enrichment
- Incident response playbooks with approval gates
- Threat intelligence summarisation
- Compliance evidence gathering
- IT service desk knowledge automation
03Across the enterprise
Enterprise
Productivity
- Document intake, classification, extraction and routing
- Executive briefing automation
- Sales and account intelligence
- Operations workflow assistance
- Training, onboarding and internal knowledge assistants
Illustrative pilot scope: a bounded document-intelligence pilot with a human-review design and no individual-impact decision or action, followed by the SOC increment.
06 / Customer-specific configuration
Six areas to tailor the platform design.
An engagement can scope agent behaviour, business knowledge, workflow, integration, governance and adoption for configuration and validation.
| Area | Examples |
|---|---|
| Domain agents | Banking BI, consumer finance intelligence, cybersecurity operations, compliance, document processing or customer support. |
| Knowledge & ontology | Approved glossaries, KPI dictionaries, metric formulas, policy libraries, report templates and prompt patterns. |
| Workflow design | Review gates, approval routing, exception handling, escalation rules and audit evidence flows. |
| Enterprise integration | Data warehouses, read-only replicas, dashboards, document stores, chat, ticketing and reporting systems. |
| Security & governance | Role-based access, masking, data boundaries, tool permissions, logging and deployment model. |
| Training & adoption | User training, administrator handover, train-the-trainer enablement and scale-out roadmap. |
Example customisations
| Use case | Target platform design | Bespoke customisation |
|---|---|---|
| FSI · Enterprise BI | Design for conversational BI agents, query validation, audit workflow and reporting templates. | English banking glossary, templates, MIS workflow deflection and role-aware data access. |
| FSI · Management intelligence | Design for management intelligence agents, KPI reporting, narrative generation and approval controls. | Consumer finance KPI ontology, daily flash briefings, NPL / DPD / collections dashboards and reporting narratives. |
| Cybersecurity SOC | Design for multi-agent triage, workflow automation, evidence requirements and approval gates. | SIEM integration, threat intelligence sources, incident playbooks and local regulatory reporting. |
| Document Intelligence | Design for document intake, extraction, classification and review routing. | Customer-specific forms, compliance rules, taxonomy and downstream workflow integration. |
07 / Commercial model
Clear components. Configure the scope.
The commercial model separates platform access, configured agents, token bundles, support and professional services.
01
Monthly platform fees
| Item | Price |
|---|---|
| dClaw IAS platform fee, up to 3 user accounts | USD 2,000 / month |
| Additional user account | USD 500 / user / month |
02
Monthly agent fees
| Item | Price |
|---|---|
| Per agent, including custom configurations | USD 1,000 / agent / month |
03
Token bundles
| Item | Price |
|---|---|
| 1,000 tokens | USD 1,000 |
04
Enterprise support
| Support tier | Scope | Price |
|---|---|---|
| Standard Support | 9x5, online chat channel | USD 1,000 / month |
| Enhanced Support | 24x7, online chat channel | USD 5,000 / month |
| Premium Support | 24x7, online chat channel, dedicated account manager | USD 10,000 / month |
05
Professional services
| Role | Rate |
|---|---|
| Senior Consultant | USD 600 / man-day (8 hours) |
| Principal Consultant | USD 1,000 / man-day (8 hours) |
Prices are reproduced from the product brief. Scope and configuration are established during solution design.
08 / How we start
A six-phase path from discovery to scale.
The proposed phases outline engagement outcomes; discovery defines what must be implemented and independently verified before any pilot use.
Discovery & solution design
Confirm use cases, users, data boundaries, governance model, deployment option and success metrics.
OUTCOMESigned-off solution blueprint
Platform setup
Provision the environment; configure roles, agents, tools, logging and support model.
OUTCOMEWorking enterprise agent foundation
Use case customisation
Configure domain agents, knowledge bases, workflows, templates, validation rules and integrations.
OUTCOMECustomer-specific working solution
Pilot & validation
Evaluate accuracy, usability, governance, audit trail, performance and support processes with pilot users.
OUTCOMEPilot evidence pack and refinement plan
Training & handover
Train users and administrators, provide playbooks, conduct handover and define a scale roadmap.
OUTCOMEOperational readiness and adoption plan
Scale & optimise
Add agents, users, workflows, integrations and dashboards based on business priority.
OUTCOMEReusable enterprise AI capability
09 / Why dClaw IAS
A reusable design for governed AI work.
The intended value is to start from a standard platform design and focus implementation on priority workflows. Outcomes depend on configuration, implementation and verification.
A standard platform design is intended to reduce one-off engineering; each engagement still requires implementation and verification.
Reusable agent, governance and integration patterns can focus implementation effort on the chosen workflow.
Designed controls can help reduce uncontrolled behaviour once they are implemented and independently verified.
A shared platform design and reusable agents are intended to avoid duplicated engineering across use cases.
The design aims to extend the same foundation through new configurations, subject to capacity and control validation.
10 / Trust & governance
Controls should be visible in the workflow.
Governance is designed into permissions, review paths and evidence capture—not left to a model's answer alone.
Human approval
The target design requires an authenticated reviewer path before release. That path is not yet built in the MVP and must be verified before pilot use.
Audit evidence
The design specifies prompt, action, tool, output, source, approval and version records. The full audit sink is not yet built; the local hash chain is not independent or immutable evidence.
Deployment choice
Customer-approved models and private-cloud, on-premise or hybrid options are design choices to scope and validate for an engagement, not claims about the current MVP deployment.
Data residency
Residency and transfer requirements need to be established in solution design. PDPC and GDPR are design targets, not certifications or a claim of achieved compliance.
Current MVP boundary
The 3-tenant, 9-user MVP is planned for one EC2 host in AWS us-west-2. Its local Ollama endpoint proxies to cloud models; a local endpoint does not mean local inference or establish data residency.
11 / Next step
Scope a bounded pilot.
Discovery can define the workflow, users, data boundaries, governance model, deployment option and success measures before implementation.
Pilot enquiries are not available through this page. Review the engagement phases to see how a pilot would be scoped.