iam-tap
Probes a local Ollama instance (/api/show, alongside llama.cpp /props and vLLM /v1/models) to discover the model's context window size, which is used to scale the middle-truncation of tool outputs to fit the context budget.
Renders the generated least-privilege IAM policy as Terraform (via --format terraform or the generate_iam_policy tool), so the emitted policy can be dropped directly into Terraform-managed AWS infrastructure code.
This server does not currently have a deployable release on Glama. Follow the repository's setup instructions to install and configure it for your MCP client, including any required credentials and local dependencies.
iam-tap
One line: npx iam-tap -- npm test (or bunx iam-tap -- bun test) produces a least-privilege IAM policy from real AWS SDK calls via Client-Side Monitoring (CSM UDP :31000).
Install
bunx iam-tap -- npm test
npx iam-tap -- npm test
# or install in your project / globally
bun add -d iam-tap
npm i -D iam-tapJSR:
bunx jsr add @kanchingyg/iam-tap
npx jsr add @kanchingyg/iam-tapgo install github.com/kanchingyg/iam-tap@latestThe core engine is Go (main.go, internal/): CSM UDP listener with token auth, argv-only child spawn (no shell), IAM policy emission, and --mcp JSON-RPC on stdio. TypeScript under src/*.ts is included for npx / bunx / JSR usage and MCP bridge helpers when no native binary is present.
Related MCP server: viberevert
Quick start
npx iam-tap -- npm test
bunx iam-tap -- bun test
npx iam-tap --format terraform -- vitest run
npx iam-tap --mcp
iam-tap --help # after go installDry-run is the default: mutating MCP tools (write_policy_file) are blocked until you pass --allow-write (which disables dry-run). Policy writes are path-contained under the working directory.
MCP Tools Reference
Tool | Mutates | Description |
|
| Spawn a child command ( |
|
| Render a least-privilege IAM policy from a recorded trace ID in |
|
| Tighten wildcard Resource ARNs using static service templates and captured region metadata. |
|
| Write the generated IAM policy to a path contained inside the current working directory (requires |
Editor and Agent Integration
Cursor (.cursor/mcp.json)
{
"mcpServers": {
"iam-tap": {
"command": "npx",
"args": ["-y", "iam-tap", "--mcp"]
}
}
}Claude Code CLI
claude mcp add iam-tap -- npx -y iam-tap --mcpProgrammatic Usage (JSR / TypeScript)
import * as bridge from "jsr:@kanchingyg/iam-tap/bridge";Runtime Safety and Guardrails
Stdio frame isolation: When
--mcpis active,hijackStdoutForMcp()redirects accidentalconsole.logcalls tostderrso JSON-RPC 2.0 frames onstdoutnever corrupt.Read-only by default: Mutating tools (
mutates: true) are blocked in--dry-runmode and require explicit--allow-writeauthorization.Loop breaker: Repeated identical
(toolName, args)invocations are hashed with SHA-256 and halted after 3 consecutive repeats to prevent runaway agent loops.Context budget truncation: Tool outputs are middle-truncated (
60%head /40%tail) scaled to the probed local modeln_ctx(Ollama/api/show, llama.cpp/props, vLLM/v1/models).Secret redaction and path normalization: AWS keys (
AKIA*,ASIA*), bearer tokens, andsk-*keys are scrubbed before returning tool output, and Windows drive paths (C:\...) translate cleanly to/mnt/c/...when WSL mode is enabled.
License
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
This server cannot be deployed
Maintenance
Related MCP Connectors
IaC attack-path auditor: finds internet-to-crown-jewel chains in Terraform/CFN/K8s.
MCP server for mandates, delegation, policy-gated execution, credential grants, and audit.
Fail-closed policy guardrails for AI agents running kubectl, terraform, helm, and argocd.
Scan prompts for injection attacks, redact PII, and audit LLM SDK usage from any MCP client
Related MCP Servers
- AlicenseNot gradedqualityDmaintenanceA task-scoped MCP stdio proxy that learns candidate least-privilege policies from labeled successful runs, requires human review, enforces exact decisions, detects tool-definition drift, and emits privacy-minimized JSONL events for Wazuh.1Apache 2.0
- AlicenseAqualityBmaintenanceLocal MCP server for AI agents and vibe coding safety: deterministic risk checks (payments, auth, database, secrets, infrastructure), sessions, checkpoints, policy inspection, and fix prompts. Runs over stdio against a local git repository; no language model judges risk.8Apache 2.0
- AlicenseNot gradedqualityBmaintenanceAggregates many MCP servers behind one local endpoint with namespaced tools, profiles, policy, budgets, audit logs, and a dashboard over stdio or HTTP.53 npm1MIT
- AlicenseNot gradedqualityBmaintenancePolicy sidecar for any MCP server: deny-by-default tool allowlist with argument rules, call and USD budget caps, and a hash-chained audit log with RFC 3161 timestamps.9Apache 2.0