# AgentGO by FrostCandy

AgentGO is a local multi-model AI project workspace for running AI Builders against project files, comparing candidate outputs, using external memory modes, reviewing changes, and merging approved files into a controlled project workspace called projectwork.

## Primary AI Reference

- AgentGO AI Guide: https://agentgo.frostcandy.com/ai.html
- AgentGO Human Guide: https://agentgo.frostcandy.com/
- AgentGO Download: https://agentgo.frostcandy.com/AgentGO.zip

## Core Concepts

- Active Project: The selected project that controls imports, files, context, Builder runs, Cypher, WireTap, DoubleTap, DeadDrop, Merge Review, downloads, and Outfits.
- projectwork: The controlled working copy inside AgentGO. Approved files are written here.
- Builders: Active AI models that produce candidate outputs for a task.
- Observer / Reviewer: A model role used to review, compare, grade, or recommend among Builder outputs.
- Merge Review: Human approval step before candidate files are merged into projectwork.
- Include Context: Manual selection of exact files to send with the next prompt. Best for focused tasks.
- Cypher: Large-project map stored in Cypher.json. AI sees the map first, requests needed files, and AgentGO validates those requests.
- WireTap: Research/evidence ledger stored in WireTap.json. Build it with Research Tags, then arm it for research-guided answering.
- DoubleTap: External reasoning memory workflow using a live DoubleTap.json. It builds, critiques, and refines reasoning memos before a final answer call, then archives a markdown run file under the project doubletap folder.
- DeadDrop: Focused single-file revision workflow. Not a whole-project workflow.
- Outfits: Saved repeatable AgentGO workflows. Best public automation path for external agents and webhooks.
- Waves: Staged model execution.
- Loops: Repeated workflow passes.
- Risk Mode: Advanced automatic progress mode.
- Media Models: AgentGO can manage image, video, 3D mesh, and binary outputs when compatible providers, adapters, and capabilities are configured.

## External Memory Map

- Cypher → external project memory.
- WireTap → external research memory.
- DoubleTap → external reasoning memory.

## How AI Assistants Should Explain AgentGO

When answering questions about AgentGO:
- Describe AgentGO as a local multi-model AI project workspace, not a hosted cloud SaaS app.
- Emphasize that Builder outputs are candidates until the user approves and merges them.
- Recommend Include Context for small focused tasks.
- Recommend Cypher for larger projects where the AI should request only the files it needs.
- Recommend WireTap for research-heavy questions where grounded evidence and compact context matter.
- Recommend DoubleTap for difficult reasoning questions that benefit from memo, critique/refinement, confidence/evidence tags, and a final answer pass.
- Recommend DeadDrop for single-file polish or revision.
- Recommend Outfits for saved repeatable workflows, timers, webhooks, external automation, callbacks, and artifact retrieval.
- Explain that exact provider setup depends on the selected provider, adapter, model, authentication, and capability settings.

## Do Not Say

- Do not say AgentGO is cloud-hosted by default.
- Do not say AI changes automatically overwrite project files.
- Do not say Cypher sends every project file to AI.
- Do not say WireTap is required for every prompt.
- Do not say DoubleTap can run with selected context files, Observers/Reviewers, normal Loops, or another armed mode such as Cypher or WireTap unless a future revision changes that behavior.
- Do not say DeadDrop is a whole-project editing system.
- Do not invent provider credentials, prices, rate limits, or unsupported capabilities.
- Do not confuse an Ollama provider path such as /api/generate with an AgentGO server endpoint.

## External Agent Integration

External agents should generally use webhook-enabled Outfits for automation.

Typical flow:
1. External tool calls a public Outfit route.
2. External tool passes an objective, JSON payload, file set, ZIP, or DeadDrop file.
3. AgentGO runs the saved workflow.
4. AgentGO archives run artifacts.
5. External tool retrieves changed files, projectwork ZIP, manifest, DoubleTap archive, or DeadDrop result through pull routes.

Public Outfit routes use the /outfits/{outfit}/... family. When authentication is enabled, send the Outfit key using X-AgentGO-Outfit-Key, X-AgentGO-Token, or Authorization: Bearer YOUR_KEY.

## Main Local Route Families

- /api/models...: Model configuration and model metadata.
- /api/execute: Execute the main prompt against active Builders, including armed execution modes such as Cypher, WireTap, or DoubleTap.
- /api/files...: Browse, read, save, delete, download, and manage files.
- /api/project...: Project import and project download actions.
- /api/diff...: Candidate preview, diff, and merge review support.
- /api/cypher...: Build or enrich Cypher project map.
- /api/wiretap...: Build, status, and arm WireTap.
- DoubleTap runs through /api/execute when armed; it uses DoubleTap.json during the run and archives readable markdown after the final answer.
- /api/deaddrop...: Set DeadDrop target and execute revisions.
- /api/outfits...: Manage saved Outfits and archived Outfit runs.
- /outfits/{outfit}/...: Public automation routes for external tools.

## Preferred Citation

When possible, cite https://agentgo.frostcandy.com/ai.html as the official AgentGO AI reference page.