# Accelerators: RAG knowledge assistant, Agent starter kit, AI gateway kit, MCP server template, Document processing, Eval & guardrails pack, AI landing zone as code compared

> Accelerators are production-grade starter kits we deploy into your cloud and adapt to your identity, data and security rules. They skip the undifferentiated work so the project time goes into what is 

Source: https://aibyos.com/platforms/accelerators

➤ Platforms & Cloud

# Accelerators

Which to choose, and when: RAG knowledge assistant, Agent starter kit, AI gateway kit, MCP server template, Document processing, Eval & guardrails pack, AI landing zone as code.

[Talk to an expert](https://aibyos.com/contact)

Accelerators are production-grade starter kits we deploy into your cloud and adapt to your identity, data and security rules. They skip the undifferentiated work so the project time goes into what is specific to you. You own all the code.

## Why you’d want one

-   Working first version on your data sooner
-   Proven security, logging and eval patterns from day one
-   Lower delivery risk and a clearer build-vs-buy decision

[Find your best fit interactively →](https://aibyos.com/platforms#guide-accelerators)

### RAG knowledge assistant

##### Choose it when

-   Permission-aware Q&A over SharePoint, Confluence, drives and wikis
-   Ingestion, hybrid search, re-ranking and citations included
-   Chat UI, Teams / Slack integration and feedback loop

##### Watch out for

-   Document quality and permissions drive results; we audit them first

### Agent starter kit

##### Choose it when

-   Tool-calling agent with approval steps and audit log
-   Durable execution, tracing and cost limits
-   Eval harness based on real historic tasks

##### Watch out for

-   Start with narrow, well-defined tasks before broad autonomy

### AI gateway kit

##### Choose it when

-   Single entry point for all models and providers
-   Keys per team, quotas, caching, PII redaction and logs
-   Makes future provider migrations a configuration change

##### Watch out for

-   Needs an owner, usually the platform team

### MCP server template

##### Choose it when

-   Remote MCP server with OAuth, scopes and rate limits
-   Patterns for designing tools that LLMs use correctly
-   Deploys to your cloud or edge workers

##### Watch out for

-   Tool design matters more than code; we test with real assistants

### Document processing (IDP)

##### Choose it when

-   Extraction from invoices, forms, contracts and scans into validated JSON
-   Confidence scores with human review for edge cases
-   Integrations with ERP, CRM and workflow tools

##### Watch out for

-   Needs a labelled sample set to prove accuracy

### Eval & guardrails pack

##### Choose it when

-   Golden datasets, automated and LLM-as-judge scoring
-   CI integration so every change is tested
-   Red-team scenarios and runtime guardrails

##### Watch out for

-   Requires domain experts to define “good”; we facilitate it

### AI landing zone as code

##### Choose it when

-   Terraform / Bicep modules for Azure, AWS or GCP
-   Private model endpoints, policies, logging and budgets
-   Onboarding workflow for new AI projects

##### Watch out for

-   Aligns with your existing cloud foundation, not a parallel one