Build withProgrammable Intelligence

The primary entry point for the SAMI ecosystem. A full-featured desktop IDE configured for repository-aware analysis and structured AI review. It handles local context indexing, secure credential management, and live multi-agent debating.

SAMI does not route your request to a single model. Every model you select sits on a council as an equal-vote Guardian for the entire run — no model outranks another. Two duties move across the council as the work advances: the Lead coordinates the discussion and announces each outcome (it has no super-vote), and the Writer is the only member that writes code and runs commands. Both duties rotate from stage to stage — they are jobs, not ranks.

Decisions are a strict binary: each Guardian votes Yes or No. A No never ends the run. It loops the council back into deliberation with the objection on the table, the members bring counter-arguments, and the work only advances once it resolves into a genuine Yes.

The same council carries a request through seven stages, from framing the problem to validating the result. Each stage scopes what its members can do: every Guardian can read and research, while writing code and running commands is reserved for the Writer and only at the two build stages. The council can run and validate the project at the end of the pipeline, surfacing a fix suggestion and re-voting when something does not hold up.

The council reasons over a shared understanding of your project, not a blank prompt. SAMI's context engine blends semantic, keyword, and code-intelligence retrieval into one project-context stream, then feeds it to the pipeline in two layers: a shared baseline context that every stage reads, plus a per-stage context slice scoped to what that stage is doing. Cross-stage recall closes the loop — any model can look back at what any other said earlier in the run, so a later stage builds on the council's own history instead of starting cold.

SAMI can connect approved external workflows to administrative and orchestration capabilities through a controlled integration layer. It speaks the Model Context Protocol (MCP), the open standard for connecting AI assistants to external tools and context — so your own assistants and automation can reach SAMI's orchestration through a single, well-defined interface rather than bespoke glue.

• MCP InteroperabilityConnect external AI assistants to SAMI over the Model Context Protocol — a standards-based bridge instead of a proprietary API surface.
• Unified Admin AccessControl users, billing, access, and system monitoring through scoped administrative actions.
• Secure Bearer AuthIssue fine-grained tokens with limited scopes to minimize attack vectors for programmatic agents.

{
  "id": "sami-01",
  "capability": "workspace_review",
  "status": "ready",
  "access": {
    "scope": "admin"
  }
}