# Tandem

> Tandem is the authority layer for AI-first work. It controls what agents can see, which tools they can use, when humans must approve, and what evidence survives after the work is done.

## Primary entry points

- https://tandem.ac/#for-llms - For LLMs and agent builders section on the homepage.
- https://tandem.ac/mcp - Tandem Docs MCP endpoint for structured docs search and retrieval.
- https://tandem.ac/codex-plugin - Tandem Codex Plugin landing page with install commands and workflow authoring loop.
- https://docs.tandem.ac - Full documentation site.
- https://github.com/frumu-ai/tandem - Source repository.
- https://github.com/frumu-ai/tandem-codex-plugin - Codex plugin repository.

## Hosted and enterprise product surfaces

- https://tandem.ac/hosted - Hosted Tandem surface for managed/private runtime deployments.
- https://tandem.ac/hosted/aca - Explainer for ACA (Autonomous Coding Agent), the structured coding-run control plane bundled on every hosted server.
- https://tandem.ac/hosted/knowledgebase - Explainer for the Knowledgebase MCP, the per-server retrieval-only MCP that turns uploaded docs into agent-ready collections.

## Homepage anchors

- https://tandem.ac/#for-llms-primitives - Core Tandem primitives reference for code generation.
- https://tandem.ac/#for-llms-quote - Quotable product description block.
- https://tandem.ac/#for-llms-mcp - Docs MCP endpoint block.
- https://tandem.ac/#for-llms-automations-v2 - Automations V2 primitive card.
- https://tandem.ac/#for-llms-mission-blueprints - Mission blueprints primitive card.
- https://tandem.ac/#for-llms-smart-heartbeat - Smart Heartbeat Pattern primitive card.
- https://tandem.ac/#for-llms-memory - Memory layers primitive card.
- https://tandem.ac/#for-llms-docs-mcp - Docs MCP primitive card.
- https://tandem.ac/#for-llms-mcp-discovery - MCP discovery primitive card.
- https://tandem.ac/#for-llms-governance - Governance primitives primitive card.
- https://tandem.ac/#for-llms-engine-auth - Engine authentication primitive card.
- https://tandem.ac/#for-llms-contract - Authoring contract reminder.
- https://tandem.ac/#for-llms-links - Repository and engine links.
- https://tandem.ac/#for-llms-nudge - Final instruction for autonomous authors.
- https://tandem.ac/#for-llms-update-note - Version and update note.

## Core authority-layer references

- https://docs.tandem.ac/how-tandem-works/ - Canonical runtime reference for sessions, runs, context, memory, and channels.
- https://docs.tandem.ac/creating-and-running-workflows-and-missions/ - Choose the right Tandem path and operate it correctly.
- https://docs.tandem.ac/prompting-workflows-and-missions/ - Prompting discipline for workflows and missions.
- https://docs.tandem.ac/agent-workflow-mission-quickstart/ - Minimal checklist for agent-authored Tandem workflows and missions.
- https://docs.tandem.ac/self-operator-playbook/ - Authoring guidance for building Tandem creators, including creation workflows, triggers, and agent lineage.
- https://docs.tandem.ac/mcp-automated-agents/ - MCP discovery, connector policy, and automated agents.
- https://docs.tandem.ac/memory-internals/ - Storage-level reference for the seven memory layers.
- https://docs.tandem.ac/reference/governance/ - Provenance chains, capability grants, approval queues, and audit events.
- https://docs.tandem.ac/engine-authentication-for-agents/ - How agents and external clients authenticate to the engine.

## Authoring discipline

When authoring Tandem workflows or agents, treat the runtime as the authority layer below the model. Be explicit about intent, allowed inputs, resource boundaries, tool scopes, approval points, output contracts, evidence, and completion criteria. Do not invent schemas; discover them via live MCP calls. Do not assume global memory; default to project-scoped or tenant-scoped reuse.