Ask
Start with the outcome, not the tool.
Describe the result. Disp8ch asks for missing details and proposes the objects to create.
Local-first AI workspace
Tell Disp8ch the outcome. It builds the work.
One request becomes workflows, decisions, tasks, and memory. On your machine, with every risky action approved by you.
local-firstapproval-gatedMIT
webchatlive
Build a daily 9am research digest and send it to WebChat. Ask before saving.
✓workflowDaily Research Digest drafted: 5 nodes, cron 9:00
✓approvalExternal send held for your confirmation
✓boardTask created: review digest sources
0workflow templates
0connected surfaces
0 keysneeded for local use
One operating loop
From question to shipped work, without changing apps.
Ground
Attach files, docs, and live app state.
Sources and notebooks keep research bounded, cited, and reusable.
Decide
Use agents when one answer is not enough.
Council records debate. Hierarchy assigns goals, roles, budgets, and owners.
Ship
Run repeatable work with visible controls.
Workflows, Boards, Activity, approvals, and Memory keep the work visible.
What it solves
Four surfaces do the heavy lifting.
Plain English control
Agentic WebChat
Ask in plain English. Approve what it proposes. Nothing mutates without you.
Explore Agentic WebChat →
Automation canvas
Visual Workflows
Describe the automation. Get an editable node graph you can test, run, and schedule.
Explore Visual Workflows →
Decision system
Council
Agents argue, vote, and record the verdict. The dissent stays on the record.
Explore Council →
Org dashboard
Hierarchy
Give agents roles, goals, and budgets. Watch heartbeats, costs, and blockers in one map.
Explore Hierarchy →Your models, your call
Local for privacy. Cloud for power. Or both.
Fully local
0 keys, 0 accounts
Point Disp8ch at Ollama, LM Studio, llama.cpp, vLLM, or SGLang. A built-in advisor recommends what fits your RAM and VRAM. Core features run offline.
Local setup →Cloud or hybrid
Bring any provider
Add a key for OpenAI, Anthropic, Google, DeepSeek, or many hosted models through OpenRouter when you want stronger results. Mix per agent and per workflow.
Model setup →FAQ
Before you install.
Do I need an API key to use Disp8ch?
No for core use. You can run a local OpenAI-compatible runtime such as Ollama, LM Studio, llama.cpp, vLLM, or SGLang. Cloud providers, live search, channels, and third-party APIs only need the credentials you choose to add.
How is Disp8ch different from a chatbot?
A chatbot answers inside one thread. Disp8ch turns plain-English requests into workflows, board tasks, council decisions, hierarchy goals, memory updates, sourced research, and saved design artifacts, with confirmation gates before risky changes.
Can Disp8ch run fully offline?
Core chat, local tools, local memory, workflows, agents, boards, Council, local document work, and local design work can run with a local model and no model-provider key. Anything that reaches the web or external services still needs network access and the credentials you configure.
Where does Disp8ch store memory?
The default workspace keeps profile and memory files on disk under data/workspace. Memory is readable, reviewable, and scoped, with review flows for learned candidates instead of silent profile rewrites.
Get started
Install it. Ask for something real.
Runs at localhost:3100. Works with a cloud key or fully local.
Linux, macOS, or WSL
curl -fsSL https://raw.githubusercontent.com/aaronnat23/disp8ch/main/scripts/install.sh | bash -s -- --repo https://github.com/aaronnat23/disp8ch.git