Choose theme
AI Resources
DeerFlow
DeerFlow is a ByteDance agent runtime for work that produces files: researching a topic, working on code, or assembling a report. It supplies skills, delegated tasks and a workspace around the model you configure.
The 2.x project is a rewrite of the original deep-research framework. Its web and terminal workflows include model-provider configuration, memory and execution tools. Those capabilities depend on the enabled tools and deployment settings. Use this as a first read, not a recommendation. Open the original project before trusting details like terms, limits, privacy, cost, setup, or safety.
What it is
Research that can produce a file
A task can involve reading sources, preparing a report and presenting the resulting document. DeerFlow supplies the workspace and tool loop around those steps; the generated result still needs checking.
Why it stands out
Files and execution belong to the workflow
Uploaded inputs, working files and final outputs have separate workspace locations. Execution can run on the host or through a container-based provider; the sandbox label alone does not describe that boundary.
Availability
Tagged releases and setup guides
The repository labels its licence MIT and publishes releases. The installation guide covers Docker and local development; model access and any connected search or messaging service require their own configuration.
Why it matters
What makes it useful
An answer in chat is not always the deliverable. DeerFlow gives an agent uploaded inputs, working files and an output area, so a research task can end with a document the reader can open. Its report skills and file-presentation tools connect the conversation to that artifact; they do not establish that the report is correct.
What to know
Where it fits
The setup wizard selects the model provider, optional search and execution settings. A lead agent can use a skill, delegate a bounded task, and bring its result back into the main workflow. A connected model and the enabled tools do the work; DeerFlow is the runtime around them, not a model checkpoint.
Notable points
What stands out
A delegated agent can receive just its task, or a snapshot of the lead agent's retained conversation and compaction summary. The README describes snapshot mode for handoffs that depend on earlier requirements or failed approaches, with more input tokens. That context choice does not change the shared filesystem, and it cannot restore messages already lost through compaction.
Before using
What to review
The installation guide requires a model configuration and credentials for the chosen provider. The README's machine-sizing notes cover DeerFlow itself; hosting a model adds separate requirements and costs.
The README recommends a trusted local network environment and additional access controls for wider deployment. Host commands, file writes and connected services are capabilities, not a safety guarantee from the word sandbox.
Choose the enabled bash, file-write, browser-use, MCP, memory and message-channel settings before sensitive work. Check what persistent memory retains and which connected channels can receive task data.
The linked documentation follows main, which can advance beyond a tagged release. The described controls need to match the installed version. LifeHubber has not run the agent or verified its isolation, privacy or generated reports.
The release page publishes v2.1.0 dated 24 September 2026, while the current security policy still says there is no official release and describes maintained branches. That inconsistency leaves tagged-release support coverage unclear; consult both records for your chosen version.
Reader fit
Who may find it relevant
Builders who want a workspace around an existing model connection, with research or coding tasks that produce files.
Operators comparing how an agent runtime handles delegated work, execution settings and persistent state.
A model download or a no-setup chatbot is a different starting point; this project requires a configured runtime and tool access.
Editorial note
Why LifeHubber lists it
DeerFlow is useful for operators maintaining shared model connections after the initial setup. Its managed catalogue provides an editing path through application Settings, with storage that can be backed up separately from the SQL database. DeerFlow's current managed-model catalogue lets operators add and edit shared OpenAI-compatible model connections in Settings without editing config.yaml. The README calls for backing up the whole managed-models directory, including its encryption key. It says this storage is independent of the SQL backend, so a database backup alone does not cover it. The README also warns that encryption does not protect against someone who can read both files.
Source links
Source materials
Reader note
Before relying on this entry
LifeHubber lists entries to help readers inspect AI projects, not to endorse them or prove they are safe, suitable, accurate, maintained, or right for a specific use. We do not verify every entry in depth. Before relying on anything listed, review the original materials, terms, privacy practices, limits, and risks that matter for your situation.
What to explore next
Compare a narrower research experiment runner.
DeerFlow covers research, coding, reports, files, memory, and sub-agents. OpenResearch CLI narrows the job to parallel code-and-compute experiments with isolated Git worktrees and visible experiment history.
More in AI Agents
Keep browsing this category
Explore more AI agent projects.
Paperclip
paperclipai/paperclip
A self-hosted server and dashboard for coordinating agent teams through companies, goals, roles, issues, heartbeats, budgets, approvals, and persistent activity records.
MiroFish
666ghj/MiroFish
A document-based multi-agent simulation project that builds personas, runs simulated social activity, and produces reports and agent conversations. Its application requires a model API and Zep Cloud; generated scenarios do not establish real-world outcomes.
Magnitude
magnitudedev/magnitude
An early open-source local inference server that profiles a computer, recommends model configurations that fit, downloads the selected model, and connects it to existing agent harnesses through local OpenAI- and Anthropic-compatible APIs.
For project maintainers
Listed here? You can use the badge.
If you maintain a project with a current LifeHubber listing, you may add the optional “Listed on LifeHubber AI Resources” badge to its README, docs, or website. No introduction or permission request is needed.
