AuraBLOCK
Cut LLM token spend with zero dependencies. AURA answers recurring prompts free from cache, compresses conversation history, and distills bloated system prompts. CLI, MCP server (Claude, Cursor, Claude Code) and library.
Overview
From the repository's own README, as read at the audited commit. Badges and raw HTML are left out.
The dependency-free context optimizer for AI agents
Part of the [⚡ SHADDAI](https://shaddai-g81x.onrender.com) family
Your agent re-sends its whole toolbox, its whole history, and its whole system prompt on every single call — most of it dead weight. AURA trims what you send before the request: sends only the tools this turn needs, compacts stale history, and distills bloated prompts. Deterministic, zero-dependency, and it fails open — it never silently drops something the model needed. Ships as a CLI, an MCP server (Claude / Cursor / Claude Code), and a library.
40-tool agent, real prompts: 82% of tool-schema tokens cut (npm run bench). At $3/M input that's ~$410 saved per 100k calls — from tool schemas alone, before history or prompt savings.[](https://www.npmjs.com/package/shaddai-aura) [](LICENSE) [](package.json) [](#-mcp-server)
**X @shaddaiAI** · **Built by @IzzoSol**
✦ Why
Every LLM app pays, again and again, for the same recurring questions. AURA intercepts them before the API call — serving deterministic answers from cache, computation, seeded facts, and parameter-less skill recipes. What can be answered for free, is.
In one line: AURA is a zero-dependency, policy-gated deterministic pre-processor for AI agents. It resolves repeated, structured, or computable prompts locally, validates reusable skills, and only lets a paid model run when you configure it to.
The core principle: keep the model off the hot path. Figure something out once, then run it deterministically
1252bb561a37OBSERVED · 2026-10-11Connect
Built from this server's own package name, version and transport as found in its source — not copied from anyone's documentation, so it cannot drift against a page we do not control. Replace the environment placeholders with a token scoped to the least it needs.
claude mcp add shaddai-aura --env ANTHROPIC_API_KEY=${ANTHROPIC_API_KEY} --env OPENAI_API_KEY=${OPENAI_API_KEY} --env OPENROUTER_API_KEY=${OPENROUTER_API_KEY} -- npx -y [email protected]{
"mcpServers": {
"shaddai-aura": {
"command": "npx",
"args": [
"-y",
"[email protected]"
],
"env": {
"ANTHROPIC_API_KEY": "${ANTHROPIC_API_KEY}",
"OPENAI_API_KEY": "${OPENAI_API_KEY}",
"OPENROUTER_API_KEY": "${OPENROUTER_API_KEY}"
}
}
}
}Exposed tools (20)
13 read · 6 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 |
|---|---|---|
aura_ask | read | Try to answer a prompt for FREE using AURA |
aura_compress | write | Shrink a conversation history BEFORE re-sending it to the model, to save tokens on every turn. |
aura_distill | write | Trim redundant instructions from a prompt or system prompt to save tokens on EVERY call. |
aura_optimize | read | One-call context optimizer: trims your tools (selective injection), distills your system prompt, and |
aura_remember | read | Teach AURA an answer so the same or a similar prompt is free next time. |
aura_savings | read | Combined AURA savings report: the answer-cache stats (prompts answered for free) PLUS the tool-cache stats |
aura_select_tools | read | Selective tool injection: given the current prompt (or recent messages) and your FULL tool list, |
aura_stats | read | Report how many tokens and dollars AURA has saved so far, and cache hit counts. |
aura_trim_output | read | Shape a tool/command result BEFORE it enters context: strips ANSI noise, collapses repeated lines, and trims |
book_flight | read | book a flight |
convert_currency | read | Convert money from one currency to another |
create_file | write | create or write a file on disk |
delete_file | destructive | delete a file from disk |
get_stock_price | read | stock quotes for a ticker |
get_weather | read | current weather and forecast for a city |
run_sql | write | run a sql query against the database |
run_sql_query | write | execute SQL against the database |
search_web | read | search the internet for information |
send_email | write | send an email |
translate_text | read | translate text between languages |
Trust audit
BLOCKgrade D · trust 67/100 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) | PASS |
| L3 | Class-specific surface | WARN |
| L4 | Behavioural (sandbox) | SKIPPED |
What the source does
- Filesystem
- declared (2 observation(s))
- Network
- declared (4 observation(s))
- Shell
- declared (1 observation(s))
- Dependencies
- pinned
- Secrets in source
- found
Findings (11)
ID.remSecret = call('tools/call', { name: 'aura_remember', arguments: { prompt: 'my stripe key', answer: 'sk_live_51Abcdefghijklmnop' } });.replace(/^\s*(what(?:'s| is)|calculate|compute|solve|eval(?:uate)?|how much is)\b/i, '')
assert.strictEqual(LS.hasSecret('my key is sk-ant-abcdefghijklmnopqrstuvwx'), true);'postgres://admin:[email protected]:5432/prod ' + 'x'.repeat(400000) // real secret + 400KB tail
['what is the github pat', 'ghp_1234567890abcdefghijklmnopqrstuv']
'Use ghp_abcdefghijklmnopqrstuvwxyz0123456 to push.'
assert.strictEqual(LS.hasSecret('token ghp_abcdefghijklmnopqrstuvwxyz0123456'), true);{ prompt: 'what is the deploy key', answer: 'GITHUB_TOKEN=ghp_abcdefghijklmnopqrstuvwxyz0123456' },delete_file
function idOf(s) { return crypto.createHash('sha1').update(salient(s).slice(0, 12).join(' ')).digest('hex').slice(0, 12); }value at cache time; secret-bearing entries (a `cat .env` result, a token in a log) stay
Gates applied: critical_finding, no_behavioural_pass.
1252bb561a37full audit observations/trust-audit/mcp-server/izzosol__aura.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 |
|---|---|---|---|---|---|
| 2026-10-11 | 1252bb561a37 | BLOCK | D | 67 | first audit |
Questions
What is the Aura MCP server?
Cut LLM token spend with zero dependencies. AURA answers recurring prompts free from cache, compresses conversation history, and distills bloated system prompts. CLI, MCP server (Claude, Cursor, Claude Code) and library.
What tools does Aura expose?
20 in total: 13 read-only, 6 that write, and 1 that can delete or overwrite (delete_file). Every one is listed on this page with its risk.
Is Aura safe to connect to an agent?
No — not without reading the findings first. The audit graded it D (67/100) and found 2 critical or high issues in the source. Each one is listed on this page with the file and line it is on. 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 Aura need?
It reads ANTHROPIC_API_KEY, OPENAI_API_KEY and OPENROUTER_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 current is this page?
The grade is for one exact copy of the source (1252bb561a37), read on 2026-10-11. The repository is watched and re-audited when it changes.