DreamGraph CLI
DreamGraph builds a project knowledge graph and serves it to agents through MCP, the dg CLI and VS Code, with ADRs.
- 128 ★ on the repository
- Last commit
Broad access
Why this level
- The daemon reads repository source and writes the graph, ADRs and schedules, and models get code fragments through your keys.
- The Computer Use backend drives a browser and requires an explicit permission policy.
Install
- Works in
Manual install
bash scripts/install.sh --forceBuilds DreamGraph, installs the `dg` CLI and the VS Code extension if `code` exists. Alternative: `npm install` and `npm run build`.
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- The daemon reads repository source and writes the graph, ADRs and schedules, and models get code fragments through your keys.
- The Computer Use backend drives a browser and requires an explicit permission policy.
This is third-party code. Review the repository files before installing.
What it does
DreamGraph runs a background daemon over stdio or HTTP that keeps a knowledge graph of a project: features, workflows, the data model, tensions, validated relationships and evidence. Agents reach it through a set of MCP tools for graph queries, enrichment, source inspection, ADRs, workflows and remediation plans. The `dg` command line creates and starts instances, attaches repositories, scans code, enriches the graph and handles schedules, export and migration. On top sits the Architect: a browser client and a terminal version, `dg architect`, for chatting with the project, plans and tool call traces. The VS Code extension adds architect chat, a dashboard and an Explorer with the graph in 2D and 3D. A cognitive engine periodically builds hypotheses ("dream cycles") and calibrates trust in them, and plugins attach through a built-in host and SDK. For multi-repo projects with a shared database there is a datastore entity and schema introspection through `DATABASE_URL`.
Who it is for. Developers and architects of large and multi-repo projects who want the agent to know how the system is built and show the grounds for its recommendations.
Good fit when
- You need to ask the agent where authorization really happens and what breaks if a module changes.
- The team keeps several repositories with a shared database and wants one knowledge graph of them.
- You want to record decisions as ADRs and see the evidence behind agent conclusions.
- You need a project chat in VS Code and an interactive dependency graph in 2D or 3D.
Not a fit when
- You need a light tool for an evening: the system includes a daemon, graph, scheduler and several clients.
- You need a license with open-source freedoms: the DreamGraph Source-Available Community License v2.0 limits terms.
- You need proven gains: a public GPT-4.1 pilot recorded quality failures and did not establish improved understanding or savings.
Example request
Create a DreamGraph instance for this repository, start the daemon and show where authorization happens in the system.Limitations
The project is maintained by a single developer. The DreamGraph Source-Available Community License v2.0 is not an open-source license. Node.js 20 or newer is required, and installation builds from source: `npm install` and `npm run build`, or the `bash scripts/install.sh --force` installer, which installs the `dg` CLI and the VS Code extension if the `code` command is found. Installing version 14 does not migrate an old graph automatically: there is a separate procedure with preview, backup and recovery. Configured OpenAI, Anthropic or Codex CLI models are needed to work. Browser access for the Computer Use backend requires an explicit permission policy, and Save and Open dialogs are supported only on Windows. A public GPT-4.1 pilot recorded quality failures.
How to disable. Stop an instance with `dg stop` and detach a project with `dg detach`, then remove the server from your MCP client. The VS Code extension is uninstalled from its menu. Instance state is kept in their directories and can be deleted.
FAQ
How do I install DreamGraph?
Install Node.js 20 or newer, clone the repository and run `npm install` and `npm run build`. Or run `bash scripts/install.sh --force`: it builds the project, installs the `dg` CLI and, if the `code` command exists, the VS Code extension.
How do I create an instance and start the daemon?
`dg init --name my-project --project /path/to/your/repo --transport http --port 8100` creates an instance, and `dg start my-project --http` starts the background daemon. For an MCP client over stdio use `dg start my-project --foreground`.
What license does DreamGraph use?
The DreamGraph Source-Available Community License v2.0: the source is available, but it is not an open-source license. Full terms are in the `LICENSE` file.
Which models are needed?
A configured link to OpenAI, Anthropic or Codex CLI models is needed, and the setup is described in `guide/04-llm-setup.md`. Model spend is billed to your own accounts.
Does upgrading to version 14 migrate an old graph?
No, installing v14 does not migrate a graph automatically. There is a separate procedure with preview, backup, cutover and recovery.
Discussion
No questions or reviews yet.
Related
CLI toolsGraphMind
graphmind
Local MCP server that builds a code graph and remembers project decisions, so the agent stops re-reading the repo every session
The /graphify skill builds a knowledge graph of a project's code, docs and PDFs so the agent queries the graph instead of grepping
Code graph and project memory in one Rust program: the agent gets excerpts with file and line citations
Structural questions about code instead of brute-forcing files: 38 languages, 42 tools, VS Code and JetBrains plugins, a CI mode
Code dependency graph for VS Code, the CLI and MCP: imports, calls, cycles, dead code and change impact.
CodeGraph (Rust)
codegraph-rust
A Rust code graph for agents: four MCP tools show what an edit will touch, how the architecture fits together and where the fragile spots are
README in short
The repository page opens with the state of version 14.0.2 and release note links, then lists the contents: daemon, MCP tools, CLI, Architect, VS Code extension, graph and cognitive engine, plugins. It continues with prerequisites, a quick start with `dg` commands, a fifteen-chapter user guide and the license terms.