Project overview
A Python and MCP-compatible context system that turns permission-scoped company knowledge into reliable working context for employees and AI assistants.
Problem
Company knowledge spread across documents, delivery records and conversations was difficult to turn into current, permission-safe working context.
Engineering scope
Company-context assembly, identity-aware access, delivery handoffs and deterministic work assurance in one internal system.
Role: Primary developer, AI context and governance engineering
Context and constraints
- Company knowledge, identities and operating rules required strict permission, ownership and confidentiality boundaries.
- Public detail is limited to the system purpose, Bongo's contribution and the technologies used; client data and operational metrics remain confidential.
Technology stack
- Python
- MCP
- JSON Schema
- YAML
- PyYAML
- GitHub Projects
- AI-assistant integration
Contributions
- Built the schema-first Python system across context assembly, source-metadata routing and deterministic work assurance.
- Designed permission-aware structured retrieval and adaptive context packs, with a read-only MCP-compatible interface for AI-assistant use.
- Implemented identity-bound handoffs, concurrent-change checks, controlled GitHub project reconciliation and permission-scoped operational dashboards.
Validation and outcome
- A deterministic verification suite covers schemas, routing, access boundaries, handoffs, assurance checks and whole-system acceptance.
- Concurrent-change and permission checks guard the transitions most likely to produce stale or unauthorized context.
Completed as a working internal AI context and control layer after a multi-month primary-development build.
Evidence scope
- Client-safe summary of implementation and validation design. Internal company records and operating metrics are not published.
Public scope
Private implementation and confidentiality boundaries limit the detail shared here. This overview focuses on the system's purpose, Bongo's role and the technologies used.
Constraints and limits
- Confidential company data, identities, operating rules and internal metrics remain outside this case study.
- The system assembles governed structured context; it is not presented as vector retrieval or autonomous decision-making.