SipcodeSAFE
Keep Claude Code context clean. Open-source toolkit: drift detection, re-read dedup, integrity scoring, AST-aware reads, 15 MCP tools. 62.6% measured savings, reproducible.
Overview
From the repository's own README, as read at the audited commit. Badges and raw HTML are left out.
Sipcode
Keep Claude Code's context clean for sharper answers and lower cost, automatically.
Website · Compare · Releases
Install in 30 seconds
npm i -g sipcode sipcode init
That is it. Your next Claude Code session will use Sipcode automatically. Verify a few minutes later:
sipcode drift
If the output reads no drift, context health stable, Sipcode is doing its job.
Already installed? Keep it fresh. Sipcode makes zero network calls, so it can't pop an "update available" notice. Build a weekly habit: runsipcode update(ornpm i -g sipcode@latest) to pull the latest fixes. Full details in [Updating Si
8d1026b04ed1OBSERVED · 2026-10-08Connect
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.
claude mcp add sipcode -- npx -y [email protected] sipcode-mcp
Exposed tools (18)
13 read · 4 write · 1 destructive. Blast radius: 1 tool can delete or overwrite — an agent that can be talked into calling a tool can be talked into calling this one.
| Tool | Risk | Description |
|---|---|---|
estimate_task_cost | read | Predict what a coding task will cost across models (Opus / Sonnet / Haiku) before the user runs it. Returns a JSON cost prediction with a confidence band and per-model token estimates. Use this when the user asks |
forecast_monthly_spend | read | Projects month-end Claude spend at the user |
get_agent_score | write | Run Sipcode |
get_drift_report | read | Detect context/cost drift: whether the user |
get_project_manifest | read | Return the Sipcode project manifest for a given directory. The manifest is a compressed <2k-token codebase map (file tree, hot files, framework fingerprint, detected patterns) generated by |
get_proxy_stats | read | Return aggregated Sipcode proxy rewrite stats: total invocations, per-rewriter counts, and estimated saved tokens (heuristic). Use when the user asks |
get_proxy_status | read | Report whether the Sipcode proxy is installed plus its accumulated rewrite stats. Read-only. Use when the user asks |
get_session_stats | read | Return cross-session token analytics (totals, per-session breakdown, top expensive sessions) as JSON, read from the local Claude Code transcripts. Use when the user asks |
get_sipcode_info | read | Return the installed Sipcode version, the list of registered MCP tools, the Node runtime version, and the host platform. Use this whenever the user asks |
get_today_summary | read | Daily dashboard — answers |
install_proxy | write | Install the Sipcode runtime proxy: writes the PreToolUse hook and registers it in ~/.claude/settings.json so Claude Code rewrites tool inputs for compact output. WRITES to settings.json (reversible with uninstall_proxy). Use when the user asks to |
list_recent_sessions | read | List the user |
sipcode_audit | write | Run a Sipcode token-usage analysis: audit the latest session, A/B before-vs-after impact, cross-session stats, or a codebase agent-friendliness score. |
sipcode_estimate | read | Predict what a coding task will cost across Claude models before running it. |
sipcode_info | read | Return Sipcode version, registered capabilities, and runtime info. |
sipcode_proxy | write | Manage the Sipcode runtime token-optimizer proxy in Claude Code: install it, uninstall it, check status, or report rewrite stats. |
uninstall_proxy | destructive | Remove the Sipcode proxy hook from ~/.claude/settings.json and delete the hook script. WRITES to settings.json. Use when the user asks to |
verify_sipcode_impact | read | Prove that Sipcode is actually saving the user tokens by A/B-comparing their token spend before vs after they installed Sipcode |
Trust audit
SAFEgrade B · trust 89/100 Nothing in the source contradicts what it says it does. Grade A is reserved for packages that have also passed the behavioural sandbox.
| Layer | What it checks | Result |
|---|---|---|
| L0 | Provenance & inventory | PASS |
| L1 | Static analysis of the code | PASS |
| L2 | Instruction surface (what it tells the agent) | PASS |
| L3 | Class-specific surface | WARN |
| L4 | Behavioural (sandbox) | SKIPPED |
What the source does
- Filesystem
- declared (2 observation(s))
- Network
- declared (1 observation(s))
- Shell
- declared (1 observation(s))
- Dependencies
- not all pinned
- Secrets in source
- none-found
Findings (15)
uninstall_proxy
.sipcodeignore
["packages/core/src/dispatcher.ts", "import { scheduleJob } from '../../workers/src/queue.js';\nexport const dispatcher = { run: (j) => scheduleJob(j) };\n"],["packages/api/src/handlers.ts", "import { dispatcher } from '../../core/src/index.js';\nexport const handlers = { run: dispatcher.run };\n"],import { dispatcher } from '../../core/src/index.js';import { scheduleJob } from '../../workers/src/queue.js';} from "../../transcript/discover.js";
expect(canonicalizeForCompare("hello")).toBe("hello");expect(canonicalizeForCompare("a\r\nb")).toBe("a\nb");const withBom = "hello world";
express
astro
@modelcontextprotocol/sdk, @resvg/resvg-js, chalk, commander, cosmiconfig, ora, prompts, satori
astro, vitest
next, react, vitest, typescript
Gates applied: no_behavioural_pass.
8d1026b04ed1full audit observations/trust-audit/mcp-server/anuj7411__sipcode.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 |
|---|---|---|---|---|---|
| 2026-10-08 | 8d1026b04ed1 | SAFE | B | 89 | first audit |
Questions
What is the Sipcode MCP server?
Keep Claude Code context clean. Open-source toolkit: drift detection, re-read dedup, integrity scoring, AST-aware reads, 15 MCP tools. 62.6% measured savings, reproducible.
What tools does Sipcode expose?
18 in total: 13 read-only, 4 that write, and 1 that can delete or overwrite (uninstall_proxy). Every one is listed on this page with its risk.
Is Sipcode safe to connect to an agent?
The audit found nothing in the source that contradicts what it says it does, and graded it B (89/100). Grade A is held back for packages that have also passed a sandboxed behavioural run, which is why a clean server reads B. Separately from the audit: 1 of its tools can destroy data, so scope the token you give it to what you actually need.
What credentials does Sipcode need?
No credential environment variables were found in its source, so it appears to need none.
How does Sipcode run?
It speaks stdio, so it runs as a local process your client starts. It is published on npm as poor at 1.6.21.
How current is this page?
The grade is for one exact copy of the source (8d1026b04ed1), read on 2026-10-08. The repository is watched and re-audited when it changes.