CladeBLOCK
Turn Claude Code and Codex from chat assistants into autonomous coding systems
Overview
From the repository's own README, as read at the audited commit. Badges and raw HTML are left out.
English | 中文
A provider-neutral delivery control plane for coding agents.
Clade turns agent work into reviewable delivery: resolved execution identity, immutable evidence, calibrated verification, correction learning, and truthful Git history. It ships native Claude Code and Codex surfaces plus an MCP bridge for other clients.
Use the full Claude Code framework, the native Codex plugin, or the provider-neutral MCP bridge. The optional Orchestrator adds a provider/model registry, evidence and eval control plane, verifier-aware routing, delivery state, and multi-project fleet truth.
If this saves you time, a star helps others find it. Something broken? Open an issue.
Blog post: Building Clade — motivation, design decisions, and lessons learned.
Table of Contents
- Install
- MCP Server
- The Trust Loop
- What It Does
- Self-Learning Mechanisms
- Skills
- Supported Languages
- Documentation
- Dotfile Sync
- Architecture
- OpenClaw Integration
- [Contributing](#co
3449d0dd95a9OBSERVED · 2026-10-10Connect
Built from this server's own package name, version and transport as found in its source — not copied from anyone's documentation, so it cannot drift against a page we do not control. Replace the environment placeholders with a token scoped to the least it needs.
claude mcp add clade-mcp -- None clade-mcp==0.3.1
Exposed tools (1)
1 read · 0 write · 0 destructive.
| Tool | Risk | Description |
|---|---|---|
clade_list_skills | read | List all available Clade skills with descriptions and argument hints. |
Trust audit
BLOCKgrade F · trust 31/100 Do not install this without reading the findings. The audit found something that could harm you or your machine.
| Layer | What it checks | Result |
|---|---|---|
| L0 | Provenance & inventory | PASS |
| L1 | Static analysis of the code | FAIL |
| L2 | Instruction surface (what it tells the agent) | FAIL |
| L3 | Class-specific surface | PASS |
| L4 | Behavioural (sandbox) | SKIPPED |
What the source does
- Filesystem
- declared (8 observation(s))
- Network
- declared (17 observation(s))
- Shell
- declared (9 observation(s))
- Dependencies
- not all pinned
- Secrets in source
- found
Findings (25)
1. **Scan the snippet for instruction-shaped patterns** (case-insensitive): `ignore previous`, `ignore prior`, `from now on`, `bypass`, `override`, `exfiltrate`, `send to https?://`, `POST to`, `webho
("SEC-01", "block", "security", re.compile(r"\beval\s*\("), "Remove eval()"),SSH_KEY="${CLADE_FLEET_KEY:-$HOME/.ssh/id_ed25519}""metadata.google.internal"]
"metadata.google.internal",
**Note:** Free sources cannot track new/lost links over time. If this section is requested without DataForSEO, inform the user: "Link velocity tracking requires the DataForSEO extension. Free sources
api = importlib.import_module(f"{module}.sync_api")mod = importlib.import_module(module_name)
"Instructs the agent to transmit data to a known exfil sink — potential exfiltration"),
OAUTH_REDIRECT_URI = "http://127.0.0.1:8085"
status, _ = _http_json(f"http://127.0.0.1:{port}/api/v1/healthz", None, timeout)f"http://127.0.0.1:{self.port}/api/v1/shutdown", token, 2.0, method="POST"f"http://127.0.0.1:{port}/api/v1/oauth/usage", token, REQUEST_TIMEOUTtext = "ANTHROPIC_API_KEY=sk-ant-api03-abcdefghij1234567890ABCDEFGHIJ1234567890XX"
text = "before sk-ant-api03-abcdefghij1234567890ABCDEFGHIJ1234567890XX after"
api_key="credential-never-serialized",
text = "token = ghp_abcdefghijklmnopqrstuvwxyz0123456789"
text = "X-API-Key: ghp_abcdefghijklmnopqrstuvwxyz0123456789"
text = "ghp_abcdefghijklmnopqrstuvwxyz0123456789"
input="ghp_abcdefghijklmnopqrstuvwxyz0123456789",
input="key=ghp_abcdefghijklmnopqrstuvwxyz0123456789",
"private_key": "-----BEGIN PRIVATE KEY-----\n...\n-----END PRIVATE KEY-----\n",
"-----BEGIN RSA PRIVATE KEY-----\n"
b"-----BEGIN PRIVATE KEY-----\nraw-private-material\n"
semantic_hash = hashlib.md5("\n".join(sorted_lines).encode()).hexdigest()[:12]Gates applied: critical_finding, instruction_override, no_behavioural_pass, undeclared_transfer.
3449d0dd95a9full audit observations/trust-audit/mcp-server/shenxingy__clade.json · Report an issue / request a re-scanAudit history
Every audit this server has had. A grade with a past is a grade somebody is still checking.
| Date | Source | Verdict | Grade | Score | Change |
|---|---|---|---|---|---|
| 2026-10-10 | 3449d0dd95a9 | BLOCK | F | 31 | first audit |
Questions
What is the Clade MCP server?
Turn Claude Code and Codex from chat assistants into autonomous coding systems
What tools does Clade expose?
1 in total: 1 read-only, 0 that write, and 0 that can delete or overwrite. Every one is listed on this page with its risk.
Is Clade safe to connect to an agent?
No — not without reading the findings first. The audit graded it F (31/100) and found 6 critical or high issues in the source. Each one is listed on this page with the file and line it is on.
What credentials does Clade need?
It reads BING_WEBMASTER_API_KEY, CLADE_API_KEY, CLADE_USAGE_HUB_TOKEN, GEMINI_API_KEY, GITHUB_TOKEN, GOOGLE_AI_API_KEY, GOOGLE_API_KEY, GOOGLE_APPLICATION_CREDENTIALS, MINIMAX_CODING_API_KEY and MOZ_API_KEY from the environment. Give it a token scoped to the least it needs — an agent that can be talked into calling a tool can be talked into calling it with your credentials.
How does Clade run?
It speaks stdio, so it runs as a local process your client starts. It is published on npm as clade-web at 0.0.0.
How current is this page?
The grade is for one exact copy of the source (3449d0dd95a9), read on 2026-10-10. The repository is watched and re-audited when it changes.