ThoughtDAG

Your thinking deserves a map
A wire is the context

Scroll

CHAT HIDES CONTEXT.
THE GRAPH IS THE CONTEXT.
Linear conversation
Editable context graph
Same prompt · different context
AI conversation 87 messages
Compare three research paths.
Start with the first. Its advantage is…
What if the core hypothesis fails?
Consider another explanation…
Also, what should I eat tonight?
There is a new restaurant nearby.

The history is here. Which parts enter the next request?

research-paper.pdfp.7

Results

The effect appears only in the experimental condition.
Selected from the page
Clipped passageresearch-paper.pdf · p.7

The effect appears only in the experimental condition

Source linked · not wired yet
Asked from sourceresearch-paper.pdf · p.7

What does this evidence actually mean?

The source is in context
Unrelated branchdetour

What should I eat tonight?

This history should not enter the research summary.

Still connected
Polluted summary 3 sources

Research summary… also, consider hot pot for dinner.

The prompt stayed the same. Polluted context changed the answer.

Includes unrelated branch
Will send1,284 tokens

Preview what the model will receive

Incoming ancestors:

Research question Evidence A Dinner detour

After deleting the orange edge: −47 tokens

Context diff−47 tok

The dinner detour left context

Same prompt · regenerate
Same promptask again

Give me a bullet-point summary

The words are identical. Only one edge changed.

Reproducible context
Clean answer 2 sources

One: record the database version. Two: use independent reviewers. Three: resolve conflicts with a third reviewer.

The unrelated dinner suggestion is gone.

Answer updated in place
One ruleThoughtDAG

Wires are context.

No hidden memory selector. What the model sees, why, and what was removed stay visible in the graph.

Visible Editable Inspectable
01 · The problem

Chat history is long. Context is still invisible.

The interface shows what was said, not which history enters the next request.

02 · Externalize

Ask from the source. Clip what matters.

Ask from a selected passage, or turn a passage or figure into its own source-linked node. Provenance stays attached; context remains yours to wire.

03 · Inspect

Before sending, inspect what the model will read.

Preview source nodes, order, and token count. Context is no longer a hidden decision.

04 · Edit

Delete one edge. Ask the same question again.

The removed branch really leaves the request. The answer changes with the context.

05 · The protocol

Most canvases organize information. ThoughtDAG edits context.

You decide what enters and leaves. The graph is the context protocol before generation.

1 / 5 Invisible context

The graph is not a picture of context. The graph is the context.

ThoughtDAG walks the incoming edges of a node, orders the relevant ancestors, and builds the message sequence sent to the selected model. The structure stays visible, editable, and inspectable.

Branch

Explore another interpretation without overwriting the path that led you here.

Prune

Keep a useful detour on the canvas while excluding it from the next request.

Merge

Bring selected evidence and reasoning paths back together in one answer.

Inspect

Preview what the model will receive before generation. No hidden memory selection.

Find the conversations.
Decide what the model sees next.

Start with a code file, an exact phrase, a URL, or a paper. ThoughtDAG finds the matching turns across your local agent conversations.

Try the CLI Read the docs

The index is built on your own machine.

Claude Code Codex DeepSeek Harness ThoughtDAG canvas
One local index
Find · one command CLI 
$ npx thoughtdag why src/lib/api.ts
why src/lib/api.ts · 12 turns in 6 sessions
claude-code · ✏ edit
  Q: Can the API detect vision support?
  Δ storedProviders → storedProviders, storedVision…
$ npx thoughtdag why src/lib/api.ts
why src/lib/api.ts · 12 个相关轮次，来自 6 个会话
claude-code · ✏ 修改
  Q: 能否判断模型是否支持多模态？
  Δ storedProviders → storedProviders, storedVision…

Do not remember the filename? An exact phrase, a URL, or a paper id can find the conversation too.

NEW IN 0.4 · SESSION ATLAS

See your agent context
as a graph

Bring work scattered across different agents into one editable graph. See what each session carries, continue from any node, then bring the new work back to where the thought began.

Local discovery Visible context Cross-agent continuity

Available in the desktop app and the Harness plugin · Claude Code, Codex and DeepSeek Harness sessions.

Choose how to open your sessions →
SESSION ATLAS context · mapped
thoughtdag-main 18 sessions
codex Investigate contradictory findings updated just now
claude-code Review benchmark design yesterday
The literature disagrees
Check study design
Alternative explanation
What stays in context?
+ new turns arrive here

NEW · INSIDE DEEPSEEK HARNESS

Your canvas. Your next harness turn.

Switch from chat to the thought graph. Connect the context, ask a follow-up, and let the harness run the next turn.

Choose DeepSeek Harness · Agent to run with tools and keep the turn in your harness session. The plugin includes the canvas; no desktop install needed.

Install the plugin →

CONTEXT REPAIR PILOT

Why editable context matters

Removing an error at its source does not mean it has left the downstream context. Across nine model endpoints, deleting only the source repaired 152 of 162 derailed cases; removing the contaminated subgraph repaired all 162. And with reasoning switched off, source-only repair collapsed to 2 of 18. In ThoughtDAG, branching, pruning and rebuilding decide what the model actually sees next.

Explore the pilot → Methods, data & results
CONTEXT REPAIR · PILOT V2 9 ENDPOINTS · 1215 CONDITIONS
Sources
Question
Explore A kept in context
Explore B outdated branch
Merge / condense next request
exclude outdated branch
152 / 162 delete source only
161 / 162 recompute downstream
162 / 162 remove whole subgraph
THOUGHTDAG PRODUCT STORY 0:32

Start with the layer you need

Search with your agent, work inside Harness, or open the standalone app. Choose one to start.

CLI + read-only MCP

Search conversations across supported agents and let your agent call the query tools directly. The desktop app is not required.

CLI and MCP manual →
npm install -g thoughtdag thoughtdag setup mcp

Already using DeepSeek Harness?

Add the thought graph to the web UI. Use your harness models, run agent turns, and explore sessions in Atlas. No separate ThoughtDAG app or CLI required.

Installation and usage → Plugin file on every release (.tgz): the same command takes its URL →
dsh plugin --profile web add dsh-thoughtdag dsh web

Full visual app · Session Atlas, canvas and file readers

The same app in its own window, local engine bundled. No Node, no terminal.

v0.3.32
macOS Apple Silicon
Install with Homebrew brew install --cask thoughtdag
Download .dmg
≈143MB
Intel Macs
Windows x64
Download installer
≈121MB
Linux x64
Download .AppImage
≈150MB
Install notes
macOS builds are signed and notarized by Apple: double-click and go. Windows builds are not signed yet; on the SmartScreen prompt choose "More info", then "Run anyway".

All versions and history live on GitHub Releases.