# Agiflow Documentation

> Your projects. Your AI. One simple board.

Agiflow is a commercial project board that connects to external AI assistants. The assistant remains the agent; Agiflow supplies scoped project-board tools, prompt skills, shared state, artifacts, vault entries, and workflow coordination.

URL: https://agiflow.ai/docs
Keywords: Agiflow, project board, ChatGPT, Claude, Cursor, scoped assistant connection, work units, tasks, artifacts, vault, workflow locks

## Getting Started

### Quickstart

Create a workspace, create a project, and connect an assistant. The recommended setup path is to copy a generated connection URL from the project board and authorize access in the browser. API keys are available for headless tools.

URL: https://agiflow.ai/docs/getting-started
Keywords: quickstart, setup, first project, assistant connection, API key

### Your First Project

Tutorial for creating a product launch project, adding tasks, connecting an assistant, verifying scope, updating task details, adding comments, and moving work through statuses.

URL: https://agiflow.ai/docs/tutorial
Keywords: tutorial, first project, marketing campaign, task comments, status update

### Connect Assistants

How-to guide for connecting ChatGPT, Claude, Cursor, or another compatible assistant to one of four direct scopes:

- Organization: `/api/v1/organizations/{organizationId}/mcp`
- Project: `/api/v1/organizations/{organizationId}/projects/{projectId}/mcp`
- Work unit: `/api/v1/organizations/{organizationId}/projects/{projectId}/work-units/{workUnitId}/mcp`
- Task: `/api/v1/organizations/{organizationId}/tasks/{taskId}/mcp`

OAuth/default resource endpoint: `/api/v1/mcp`.

URL: https://agiflow.ai/docs/connecting-ai-tools
Keywords: ChatGPT, Claude, Cursor, OAuth, API key, organization scope, project scope, work unit scope, task scope

## Reference

### Capability Reference

Agiflow platform sessions expose three progressive-discovery meta-tools:

- `describe_capabilities`
- `use_capability`
- `get_current_scope`

Core tool families include projects, work units, tasks, task comments, members, organizations, artifacts, vault entries, and workflow locks.

Artifact tools: `list_artifacts`, `get_artifact_signed_url`, `update_artifact`, `link_artifact`, `delete_artifact`.

Vault tools: `list_vault_entries`, `get_vault_entry`, `set_vault_entry`, `delete_vault_entry`, `bulk_upsert_vault_entries`.

Workflow lock tools: `acquire-workflow-lock`, `release-workflow-lock`, `get-workflow`, `list-workflows`, `update-workflow`, `delete-workflow`, `check-workflow-lock`.

Prompt skills: `getting-started`, `project-plan`, `run-work`, `run-task`, `refine-task`, `backlog-grooming`, `review-work`, `triage`, `daily-standup`, `orchestrate`.

URL: https://agiflow.ai/docs/features/ai-skills
Keywords: capabilities, prompt skills, artifacts, vault, workflow locks, work units, widgets

## Explanations

### Project Workflows

Agiflow models work with projects, work units, tasks, statuses, artifacts, vault environments, and workflow locks. External assistants can help plan, refine, run, review, and report work by calling scoped tools. The project board remains the source of truth.

URL: https://agiflow.ai/docs/features/workflows
Keywords: project workflow, work unit, task status, artifact, vault environment, workflow lock

### Security & Governance

Access is layered through Better Auth user authentication, organization membership, project access middleware, admin route guards, scoped assistant authorization, and API key validation. Auth data lives in D1, project data in Durable Object SQLite, and artifact blobs in R2. Vault entries are encrypted at rest with per-entry IVs and masked in list/get responses. Plan quotas are enforced at API ingress.

URL: https://agiflow.ai/docs/features/security
Keywords: security, Better Auth, API key, OAuth, project access, Durable Object, D1, R2, vault encryption, quota

## Guides

- How your board is organized: https://agiflow.ai/guides/project-organization
  Explains the three-level model (Projects, Work Units, Tasks), the eight default board statuses, built-in templates, project artifacts, and team access.

## How-To Guides

- Start from a project template: https://agiflow.ai/guides/project-templates/use-template
- Run a marketing campaign: https://agiflow.ai/docs/guides/marketing-campaign
- Track sales deals: https://agiflow.ai/docs/guides/sales-tracking
- Manage freelance projects: https://agiflow.ai/docs/guides/freelancer-projects

## Competitor Comparisons

### Jira Alternative

A detailed comparison of Agiflow versus Jira for small technical teams using AI assistants. Highlights how Agiflow provides focused kanban boards, artifacts, vaults, and workflow locks via MCP tools, while acknowledging Jira's strength in sprints, advanced roadmaps, and enterprise ceremonies.

URL: https://agiflow.ai/jira-alternative/
Keywords: Jira alternative, project management comparison, MCP AI task tracking, developer board

### Trello Alternative

A comparison of Agiflow versus Trello for teams that like a simple kanban board but want more structure and a board their AI assistant can use. Agiflow keeps cards, columns, and drag-and-drop, adds projects and work units above flat cards, and lets ChatGPT, Claude, and Cursor read and update the live board. Acknowledges Trello's strengths in its 200+ Power-Up marketplace, Butler automation, advanced Premium views, and mature mobile app. Trello pricing referenced as of May 2026.

URL: https://agiflow.ai/trello-alternative/
Keywords: Trello alternative, Trello alternative for AI, kanban board, AI project management, ChatGPT project board, flat per-seat pricing

### Asana Alternative

A detailed comparison of Agiflow versus Asana for small and technical teams that want their own AI assistants connected over MCP. Highlights how Agiflow provides a focused kanban board, project artifacts, per-environment vaults, and workflow locks, while acknowledging Asana ships an official MCP server and remains stronger for built-in AI agents, goals and portfolios, broad view types, and enterprise governance. Asana figures are as of May 2026.

URL: https://agiflow.ai/asana-alternative/
Keywords: Asana alternative, project management comparison, MCP AI project management, bring-your-own-assistant board

## Integrations

Agiflow publishes per-client integration pages that explain how to connect a specific AI client or coding agent to an Agiflow board over the Model Context Protocol (MCP). The integrations hub groups them into AI clients (ChatGPT, Claude, and any MCP-compatible client) and Agents (Cursor, Codex, Claude Code, and Antigravity).

URL: https://agiflow.ai/integrations/
Keywords: Agiflow integrations, AI client integration, coding agents, MCP integration, integration directory

### Claude Integration

Connect Claude (Anthropic) to Agiflow as a custom MCP connector so Claude can read approved projects and work units, create and update tasks, and add task comments on the board you authorize. Covers setup for Claude.ai web and Claude Desktop. Connector URL: https://agiflow.io/api/v1/mcp-app/0.0.1. For Claude Code terminal setup, use the dedicated Claude Code integration.

URL: https://agiflow.io/integrations/claude/
Keywords: Claude integration, Claude Agiflow integration, Claude connector for project management, Claude MCP connector

### ChatGPT Integration

Connect ChatGPT to Agiflow so the assistant works from the project board you authorize. The integration connects ChatGPT to Agiflow's MCP server through the published ChatGPT app: you connect from the ChatGPT app listing, approve access, and choose the workspace, project, or task ChatGPT can use. ChatGPT can then review project context and create or update tasks, work units, and comments within that scope. Agiflow provides tools and widgets to ChatGPT and does not run agents itself.

URL: https://agiflow.ai/integrations/chatgpt/
Keywords: ChatGPT integration, Agiflow ChatGPT app, ChatGPT project management, AI client integration, MCP project management

### Claude Code Integration

Connect Anthropic's Claude Code terminal coding agent to Agiflow over MCP so it can read and update the projects, tasks, work units, workflows, comments, members, and vault entries on your board, and run Agiflow's workflow skills — instead of working from a pasted brief. Two setup paths: the Agiflow AI Plugin (`claude --plugin-dir ./agiflow-ai-plugin`, which auto-wires the MCP server from the bundled `.mcp.json`) or a direct connection (`claude mcp add --transport http agiflow YOUR_AGIFLOW_URL`, then `/mcp` → Authenticate). Coding-agent MCP endpoint: `https://agiflow.io/api/v1/mcp` (not `/api/v1/mcp-app`, which serves the ChatGPT widget MCP).

URL: https://agiflow.ai/integrations/claude-code/
Keywords: Claude Code integration, Claude Code Agiflow, Anthropic coding agent, Claude Code MCP, Agiflow AI Plugin, claude mcp add, terminal coding agent

### Cursor Integration

Connect Cursor (the AI code editor by Anysphere) to your Agiflow board over MCP so the editor can read and update Agiflow projects, work units, tasks, and task comments, scoped by the connection URL you authorize. Setup: copy your Agiflow connection URL from Settings → Connections, add it to `.cursor/mcp.json` as `YOUR_AGIFLOW_URL`, restart Cursor, then enable Agiflow and authorize in the browser.

URL: https://agiflow.ai/integrations/cursor/
Keywords: Cursor integration, Cursor MCP, connect Cursor to Agiflow, coding agent, .cursor/mcp.json, scoped connection URL

### Codex Integration

Connect OpenAI Codex — the coding agent across the Codex CLI, IDE extension, and cloud — to your Agiflow board over MCP so it can read and update Agiflow projects, work units, tasks, and task comments, scoped by the connection URL you authorize. Setup: install or open Codex, copy your scoped Agiflow connection URL from Settings → Connections, add a `[mcp_servers.agiflow]` table with `url = "YOUR_AGIFLOW_URL"` to `~/.codex/config.toml` (or a trusted project `.codex/config.toml`), then run `codex mcp login agiflow` and verify with `codex mcp list`. Coding-agent MCP endpoint: `https://agiflow.io/api/v1/mcp` (not `/api/v1/mcp-app/0.0.1`, which serves the ChatGPT app/widget MCP surface).

URL: https://agiflow.ai/integrations/codex/
Keywords: Codex integration, OpenAI Codex MCP, connect Codex to Agiflow, codex config.toml mcp, codex mcp login, coding agent project management

### Antigravity Integration

Connect Google Antigravity — an agentic development platform / coding agent — to Agiflow by installing the first-party Agiflow AI Plugin as a local Antigravity plugin directory. After install and restart, Antigravity can read and update Agiflow projects, work units, tasks, comments, artifacts, and workflow context, and load the plugin's bundled workflow skills — scoped to the resource the connected user authorizes. Setup: clone https://github.com/AgiFlow/ai-plugin, copy it into `.agents/plugins/` (or `_agents/plugins/`) for one workspace or `~/.gemini/config/plugins/` globally, restart Antigravity so it auto-reads `plugin.json`, `skills/`, and `mcp_config.json`, then approve the Agiflow connection. The plugin bundles `serverUrl: https://agiflow.io/api/v1/mcp` with header `x-agiflow-mcp-consumer: plugin`; the remote key is `serverUrl`, not `url`; Antigravity's manual config file is `~/.gemini/config/mcp_config.json`; self-hosted deployments override the endpoint with `AGIFLOW_AI_PLUGIN_MCP_URL`. There is no public Antigravity marketplace listing for Agiflow yet — install as a local plugin directory. Agiflow provides MCP tools and does not run agents or imply Google partnership, certification, or endorsement.

URL: https://agiflow.ai/integrations/antigravity/
Keywords: Antigravity integration, Google Antigravity, Agiflow AI Plugin, antigravity mcp server, connect Antigravity to Agiflow, coding agent project management, serverUrl, ~/.gemini/config/plugins

## Open Source Toolkit

The open-source component is aicode-toolkit, not the Agiflow product/API/app.

- Repository: https://github.com/AgiFlow/aicode-toolkit
- Scaffold MCP: https://agiflow.ai/docs/mcps/scaffold-mcp
- Architect MCP / vibe-lint: https://agiflow.ai/docs/mcps/architect-mcp
- One MCP: https://agiflow.ai/docs/mcps/one-mcp

One MCP toolkit meta-tools are `describe_tools` and `use_tool`. Agiflow platform meta-tools are `describe_capabilities`, `use_capability`, and `get_current_scope`.