Scan your whole stack
Latest version: v0.15.0
redthread was last scanned on 18 Nov 2026 at version 0.15.0. Risk score 0 of 100. No malicious behavior was detected.
| Version | Published | Status |
|---|---|---|
| 0.15.0 LATEST | 18 Nov 2026 | No rules fired |
Findings for v0.15.0. The score is a capped sum of the weights of the rules that fired — not CVSS, not exploitability, and not a probability. It is not comparable across packages with different scan coverage.
The score describes 0.15.0; other versions have their own rules, not their own score.
The publisher’s own README, as collected from the registry.
Portable, git-backed memory for AI agents and multi-phase workflows.
A coding agent usually keeps its memory in a local folder (.claude/,
.agent/) on one machine. Redthread replaces this folder with a
git-backed memory store that an MCP server gives to the agent. Each
machine that clones the store shows the same memory to the agent.
The same store also keeps the context of a multi-phase pipeline, for example train → eval → present or build → test → present. This context is an append-only, content-addressed memory. A git remote is its source of truth. Any node can clone the store and continue a run. Each phase publishes a curated handoff to the subsequent phase.
Phase names are data, not code. Each project declares its own pipeline in
project.yaml. Code below the adapter layer has no domain-specific terms.
.claude/.agent folder.mcp-serve without
--store. Each workspace then gets the store that its committed
.redthread.yaml file names. Thus, one global MCP entry (Claude Code
user scope, Cursor, Windsurf, VS Code) operates correctly in all your
repos. In a repo that has no store yet, the agent's store_init call
makes one as an orphan-branch worktree, with no CLI step.project_id /
run_id / phase / entry_id. The key is never a hostname or an
absolute path. To move a phase to a different machine during a run, use
redthread resume.redthread present makes a
markdown report, a slide deck, and a docs-site tree from a run.git worktree of
your code repo. Redthread does not change the active branch of that repo..redthread.yaml
marker lets redthread mcp-serve and redthread attach find and attach
the store automatically. On a new clone, only git clone and the same
MCP command are necessary. You do not have to remember flags.pip install redthread # or: uv tool install redthread
Copy the AGENTS.md example into your project. The next time an agent opens the project, it uses Redthread as its memory:
store_init.By default, the store is an orphan-branch git worktree of the same repo. Thus, you do not have to set up a second remote, and the active branch of the repo does not change.
Register the server one time. You do not make a store first, and you do not give a
--store path: the same registration serves every project. The first time an agent
uses Redthread in a repo, it asks you, then calls store_init, which makes the store
as an orphan-branch worktree at ./redthread-store. Select your client:
claude mcp add --scope user redthread -- uvx redthread mcp-serve
If redthread is already installed, remove uvx:
claude mcp add --scope user redthread -- redthread mcp-serve
To make sure that the server operates, type /mcp in Claude Code. The
redthread server shows as connected with 19 tools.
Cursor does not install MCP servers with a CLI command. It uses a one-click deeplink. This command makes the deeplink and opens it. It uses only Python, which Redthread already uses:
python -c "
import base64, json, webbrowser
config = {'command': 'uvx', 'args': ['redthread', 'mcp-serve']}
encoded = base64.b64encode(json.dumps(config).encode()).decode()
webbrowser.open(f'cursor://anysphere.cursor-deeplink/mcp/install?name=redthread&config={encoded}')
"
Cursor shows an install confirmation. Accept it to complete the procedure.
code --add-mcp '{"name":"redthread","command":"uvx","args":["redthread","mcp-serve"]}'
If you use the Insiders build, use code-insiders instead of code.
You can also connect Windsurf, Claude Desktop, Codex CLI, Gemini CLI, and the Claude Agent SDK. For each client, refer to MCP client setup.
After you connect the agent, tell it to call agents_md_bootstrap. This
tool writes the usage policy from Option 1 into the AGENTS.md/CLAUDE.md
file of the project. Future sessions then use this memory automatically.
redthread run start --store ./my-store
redthread log <run_id> build note '{"msg": "hello"}' --store ./my-store
redthread read <run_id> --store ./my-store
If you use a source checkout, run uv sync first. Then put uv run before
each command.
The full docs (usage, architecture, store format) are at
sina5.github.io/redthread. The
source is in docs/. MkDocs Material builds the site. To show the docs
locally, run this command:
uv run --group docs mkdocs serve
For release notes, refer to CHANGELOG.md.
uv sync
uv run pytest
uv run ruff check . Find threats in your stack.
OX Radar can check high-priority threats, like critical package CVEs, against your context.
Each incident is marked Detected or Not Detected with a remediation plan and proof all tracked in the Response Center.
The probe returned only these fields for this server — auth requirement, protocol and tool list weren’t observed, so they’re not shown (never guessed).
The publisher’s own links, as the registry reports them. These are not sources for the assessment on this page.