This guide is for people using AgentStack through the website (agentstack.tech): sign in, pick a project, and work in the dashboard. It summarizes recent capabilities (RAG, sandboxes, access control) in plain language and points to deeper docs when you need API or automation details.
Language: This public documentation is English only.
After you log in, open Dashboard and select a project. Modules you may see include:
| Module | What you use it for |
|---|---|
| Overview | Project snapshot and entry to other areas. |
| Project settings | Name, type, and project-level options. |
| Sessions | Active sessions tied to your project. |
| RBAC | Roles and permissions for people on the project. |
| Field access | Who can see which fields in API responses (masking / hiding sensitive data). |
| RAG | Knowledge collections, document ingest, semantic search, and optional session memory for AI context. |
| Buffs | Trials, subscriptions, and promotional entitlements when enabled for your project. |
Exact labels depend on your subscription and permissions. If a module is missing, your role or plan may not include it.
URL pattern (examples):
https://agentstack.tech/dashboard/<projectId>?module=rag — opens the dashboard with the RAG module selected (when available).
RAG (Retrieval-Augmented Generation) lets you attach searchable knowledge to a project: upload text, search by meaning, and (where enabled) keep a short memory of a conversation session so AI tools can stay on topic.
Typical workflow in the UI
- Open Dashboard → RAG for your project.
- Create a collection (a bucket for related documents).
- Add documents (text is chunked and indexed automatically).
- Run semantic search inside a collection.
- Optionally use session memory features if your workflow needs recent turns indexed for retrieval.
Limits depend on your subscription (collections, chunks, memory rows, etc.). If an action is blocked, check your plan or upgrade path on the site.
More detail (includes REST /api/rag/* and MCP for automations):
RAG_PLATFORM_GUIDE.md
Sandboxes let you work in an isolated environment derived from your project: try changes, checkpoints, and (on supported plans) promotion flows back toward production — without replacing your live data blindly.
This is aimed at teams that need safe experimentation and controlled rollout. Day-to-day users may only see sandbox features if the product surface exposes them for your account.
Deeper technical guides (APIs, limits, flows):
- SANDBOX_PLAYGROUND_GUIDE.md — concepts, subscription gates, and patterns.
- SANDBOX_AND_ENVIRONMENTS.md — REST API and environment behaviour.
API reference: Swagger UI — look for Sandbox-related tags and OPENAPI.md.
AgentStack uses layered access:
- API keys — may be restricted to certain services (automation safety).
- Roles (RBAC) — who may open which parts of the product and call which APIs.
- Field Access Policy (FAP) — which fields appear in JSON for each role (read, mask, or hide).
As a project owner or admin, you may configure field access in the dashboard. As a member, you simply see what your role allows.
Readable overview: ACCESS_AND_FIELD_POLICY.md
Full policy format: FIELD_ACCESS_POLICY.md
Plans differ by projects, members, API usage, storage, RAG, sandbox features, support, and more.
Comparison table: subscription/SUBSCRIPTION_TIERS.md
Anonymous trial-style usage: subscription/ANONYMOUS_TIER.md
Pricing and billing UI: use agentstack.tech (Pricing / Billing). For raw API details, use Swagger.
The AI Builder may be presented as a demo of the ecosystem or temporarily unavailable on the public site. When it is off, use the rest of the dashboard, APIs, and MCP integrations for your workflows.
If you use MCP or plugins, start from the ecosystem index and the plugins list:
| Topic | Document |
|---|---|
| Full doc index | README.md |
| MCP + ecosystem index | MCP_AND_ECOSYSTEM.md |
| OpenAPI / Swagger | OPENAPI.md |
| REST API topics | api/README.md |