Install
$ agentstack add mcp-axis-love-flow ✓ scanned · ✓ verified, works with Claude Code, Cursor, and more.
Security review
✓ PassedNo issues found. Passed automated security review. · v0.1.0 How review works →
- ✓ Prompt-injection patterns
- ✓ Secret / credential exfiltration
- ✓ Dangerous shell & filesystem operations
- ✓ Untrusted network calls
- ✓ Known-malicious package signatures
What it can access
- ● Network access Used
- ✓ Filesystem access No
- ✓ Shell / process execution No
- ✓ Environment & secrets No
- ✓ Dynamic code execution No
From automated source analysis of v0.1.0. “Used” means the capability is present in the source — more access means more to trust, not that it’s unsafe.
Verified badge
Passed review? Show it. Paste this badge into your README, it links to the public security report.
Reliability & compatibility
Declared compatibility
Compatibility is declared by the source manifest. End-to-end runtime verification is coming, see below.
We're building live execution health for every listing: tool-call success rate, median latency, uptime, and last-checked timestamps, measured, not self-reported. It isn't live yet, so we don't show numbers we can't stand behind.
How agent discovery & health will work →About
Agent-first Kanban board — self-contained, SQLite-backed, no database server
[](LICENSE)
Why Flow?
Task boards built for humans assume a person is always in the loop. Flow doesn't. It treats agents as first-class users — role-scoped API keys, an MCP endpoint, and human-required flags let LLM agents pick up work, move tasks, and signal blockers without stepping on each other or silently drifting past decisions that need a human.
Single FastAPI process backed by SQLite. No database server, no external auth, no message broker. Deploy as a binary, a container, or just uvicorn flow_app.main:app.
Features
- Kanban board — five columns with enforced role transitions
- Dependency graph — hover any card to see parent/child lines across columns
- Role-scoped API keys — admin, architect, implementer, reviewer, read_only
- Human-required flag — tasks that block on human decisions, with reason
- MCP interface — JSON-RPC 2.0 endpoint for LLM agents (
POST /mcp) - Ideas intake — capture ideas, promote them into linked tasks
- Markdown import — preview and commit tasks from Markdown files
- Qualification fields — complexity, impact, effort, risk on every task
- Two themes — Neutral (default) and Axis Love
- Single-binary deployment — FastAPI + SQLite, zero external dependencies
Quick Start
pip install . && flow-serve --bootstrap
flow-serve --bootstrap seeds the default project and prints role-scoped API keys once — copy them. It starts the board on http://localhost:8100 (host/port from FLOW_HOST/FLOW_PORT). To use the board UI, click Sign in and paste an admin key. (Or just run ./scripts/quickstart.sh, which does all three steps and prints the connect instructions.)
Docker
# Web service only
docker compose up -d
docker compose logs flow # first run prints admin / implementer / reviewer keys once
# Web service + automation runner
docker compose --profile runner up -d
> Deploying to a server? See the [server deployment checklist](docs/Operations.md#deploying-to-a-server) for the HTTPS, session-cookie, and proxy settings that differ from a local run.
Signing in
flow-serve --bootstrap prints an admin key on first run. Open http://localhost:8100, click Sign in, and paste it — you now have a browser session and can manage tasks, ideas, and more keys under Settings → API keys. (No reverse proxy or extra config needed; flow-serve sets up the session secret automatically.)
The same keys work for the API:
curl -H "Authorization: Bearer YOUR_KEY" \
-H "Content-Type: application/json" \
http://localhost:8100/api/tasks
For agents
Flow is built for LLM agents, and onboarding one is a single link:
- Working in this repo? Point the agent at [AGENTS.md](AGENTS.md) — one-command setup, how to
connect via MCP, and the core work loop.
- Connecting to a running server? Hand it
http://:8100/llms.txt. That endpoint
self-describes the live deployment (its MCP URL, a ready-to-paste claude mcp add command, and the agent loop) so the agent can configure itself.
Claude Code, in one command:
claude mcp add --transport http flow http://localhost:8100/mcp \
--header "Authorization: Bearer YOUR_KEY"
Hermes and other clients: install the skill at [docs/FLOW-HERMES-SKILL.md](docs/FLOW-HERMES-SKILL.md).
Documentation
All docs live in [docs/](docs/) — start at [docs/README.md](docs/README.md) for the full index.
| Document | Description | |----------|-------------| | [Agent Quickstart](docs/AGENT-QUICKSTART.md) | Connect an agent and run the core work loop | | [Architecture](docs/Architecture.md) | System design, data model, lifecycle | | [Operations](docs/Operations.md) | Setup, deployment, backup | | [REST API](docs/Modules/REST-API.md) | Complete REST API reference | | [MCP](docs/Modules/MCP.md) | MCP interface for LLM agents | | [Security](docs/Modules/Security.md) | Roles, permissions, API keys | | [Web UI](docs/Modules/Web-UI.md) | Board, themes, overlays |
Architecture
┌─────────────────────────────────────────┐
│ FastAPI Application (flow_app.main) │
│ ┌───────────┬───────────┬───────────┐ │
│ │ REST API │ MCP API │ HTML UI │ │
│ └─────┬─────┴─────┬─────┴─────┬─────┘ │
│ │ │ │ │
│ ┌─────┴───────────┴───────────┴─────┐ │
│ │ Repository Layer │ │
│ └─────────────────┬─────────────────┘ │
│ │ │
│ ┌─────────────────┴─────────────────┐ │
│ │ SQLAlchemy + SQLite │ │
│ └───────────────────────────────────┘ │
└─────────────────────────────────────────┘
Contributing
Contributions welcome! See [CONTRIBUTING.md](CONTRIBUTING.md) for guidelines.
Good first issues: [good first issue](../../issues?q=is%3Aissue+is%3Aopen+label%3A%22good+first+issue%22) · [help wanted](../../issues?q=is%3Aissue+is%3Aopen+label%3A%22help+wanted%22)
License
[MIT](LICENSE) © Nyx Prime
Source & license
This open-source MCP server is cataloged on AgentStack and links to its original source — we do not rehost the code.
- Author: axis-love
- Source: axis-love/flow
- License: MIT
Install and usage instructions live in the source repository linked above.
Reviews
No reviews yet, be the first.
Write a review
Versions
- v0.1.0 Imported from the upstream source.