AgenticLedger
Local-first flight recorder for AI agents: every LLM call, tool call, cost, and loop captured by a transparent proxy. Zero code changes.
Emerging. No concerning findings. Grades remain capped until the project builds reputation through adoption. Grades reflect the full trust pyramid: code, verification depth, and reputation. New projects cap at C until adoption is earned.
How we verified
Verified is a snapshot. Live keeps it current, and builds your track record.
⚡ Connect GitHub → continuous verification on every pushwhy connect →Who stands behind it
ShekharBhardwaj
Source: Glama · also listed on github_repo_search
Claim it to get a verified publisher badge, a free copy of our full audit findings, and direct contact for any high-priority issues we find. Or connect your repo for our deepest verification, Live Monitored: read-only, revoke anytime. What we access →
Install from
The grade above is for the source repository. Registries can serve a different version, so we mark the ones we were not able to read.
SMOKE_DBSMOKE_PYTHONAGENTICLEDGER_OTEL_ENDPOINTOpenTelemetry? Install the extra and set :AGENTICLEDGER_OTEL_HEADERSComma-separated key=value auth headers, e.g. x-honeycomb-team=abc123.AGENTICLEDGER_OTEL_SERVICE_NAMEValue of service.name reported to the collector.AGENTICLEDGER_UPSTREAM_URLLLM endpoint to forward requests to. Accepts OpenAI, Anthropic, LiteLLM, OpenRouter, or any OpenAI-compatible URL. Omit it and the proxy routes each call by its wire format: Anthropic-shaped calls to Anthropic, Bedrock paths to Bedrock, everything else to OpenAI.AGENTICLEDGER_HOSTHost to bind to. Use 127.0.0.1 to restrict to localhost only.AGENTICLEDGER_PORTper-setting (=9000 agenticledger start uses 9000AGENTICLEDGER_BUDGET_STATUSHTTP status for budget blocks. 429 ships with an honest Retry-After (seconds until the UTC-midnight window reset); set 402 if your clients retry 429s aggressively — nothing retries Payment Required.AGENTICLEDGER_BUDGET_ACTIONWhat happens when a budget is exceeded: block returns HTTP 429 (call never reaches the LLM), warn lets the call through and fires a webhook alert, both blocks and fires the webhook.AGENTICLEDGER_RATE_LIMIT_RPMMax requests per minute globally.AGENTICLEDGER_RATE_LIMIT_SESSION_RPMMax requests per minute per session_id.AGENTICLEDGER_RATE_LIMIT_AGENT_RPMMax requests per minute per agent_name.AGENTICLEDGER_RATE_LIMIT_USER_RPMMax requests per minute per user_id.AGENTICLEDGER_ALERT_WEBHOOK_URLURL to POST alert payloads to. Required for any alerts to fire.AGENTICLEDGER_ASYNC_CAPTUREPersist captures on a background worker so storage never adds latency to the agent's call. Trade-off: reads become eventually consistent (a just-captured call may not be queryable for a brief moment). Recommended for high throughput.AGENTICLEDGER_CAPTURE_QUEUE_MAXMax captures buffered in async mode before load is shed (drops are counted in /metrics).AGENTICLEDGER_CAPTURE_LEVELfull stores everything; metadata stores only metrics/metadata (model, tokens, cost, latency, agent, status) and drops prompts, responses, and tools.AGENTICLEDGER_REDACTRedact PII/secrets in stored data: all, or a comma list of email,ssn,credit_card,ip,api_key. Replaces matches with [REDACTED:<label>]. Only the stored copy is affected — the agent's response is untouched.AGENTICLEDGER_REDACT_PATTERNSExtra redaction regexes as JSON: {"label": "regex", ...} or ["regex", ...].AGENTICLEDGER_AUDIT_LOGRecord an audit trail of who viewed/exported/deleted what plus token/erasure actions. Set 0 to disable.AGENTICLEDGER_LOOP_ACTIONwarn records loop_flags and fires a loop_flag webhook alert; block additionally returns HTTP 429 (loop_detected) for a session that tripped a guard; off disables inference.AGENTICLEDGER_LOOP_MAX_STEPSFlag (and in block mode, stop) threads that exceed this many ReAct steps.AGENTICLEDGER_LOOP_REPEAT_THRESHOLDConsecutive identical tool calls (same tool, same arguments) before a thread is flagged stuck.AGENTICLEDGER_LOOP_RUN_GAP_SECONDSMax gap between fresh-context spawns (same system prompt) that still count as iterations of one run.AGENTICLEDGER_COMPLETION_PROMISERegex matched against response text. On match the call is flagged completion_promise and the run's status becomes complete — loop runners poll GET /api/runs/{run_id} and stop.AGENTICLEDGER_DIGEST_HOURUTC hour (0–23) to POST a daily spend digest — last 24h totals, cache savings, top models/agents — to the alert webhook. Slack-incoming-webhook friendly (text).AGENTICLEDGER_EXTRA_PATHSComma-separated additional request paths to capture, e.g. v1/responses,v1/custom. Built-in paths (v1/chat/completions, v1/messages, v1/responses, plus v1/messages/count_tokens recorded as a free call) are always captured.AGENTICLEDGER_INSTANCEAGENTICLEDGER_API_KEYIf is set, pass it as a header:AGENTICLEDGER_URLAGENTICLEDGER_CONFIG1. The file is found in this order: , thenWIRETAP_TAGAWS_REGIONAWS_DEFAULT_REGIONAGENTICLEDGER_EXPORT_HMAC_KEYWhen set, compliance exports carry a tamper-evident keyed hmac-sha256 integrity tag instead of a plain sha256 checksum.AGENTICLEDGER_PRICINGInline JSON map of model → [input_per_million, output_per_million] USD. E.g. '{"gpt-4o": [2.50, 10.00], "my-model": [1.00, 2.00]}'.AGENTICLEDGER_PRICING_FILEPath to a JSON file with the same format. Applied after AGENTICLEDGER_PRICING.AGENTICLEDGER_DSNTool annotations
No tools have read-only/destructive annotations
Add readOnlyHint or destructiveHint annotations to every tool so hosts can warn users before invoking.
All four hints declared on every tool
6/6 tools missing one or more hints — list_sessions (missing: readOnlyHint, destructiveHint, idempotentHint, openWorldHint); explain (missing: readOnlyHint, destructiveHint, idempotentHint, openWorldHint); get_session (missing: readOnlyHint, destructiveHint, idempotentHint, openWorldHint), +3 more. OpenAI's directory rejects tools where any of the four hints are missing or non-boolean.
For every tool, set all four hints (readOnlyHint, destructiveHint, idempotentHint, openWorldHint) to explicit true/false values that match the handler’s actual behaviour.
Secrets not logged
1 secret value sent to print
Redact or omit secret values from log output.
Claim the listing to review these findings one by one and send us a correction where you disagree, straight to the team. Claiming also means we tell you when the grade moves, and reach you first if we find anything urgent.
[](https://m8ven.ai/mcp/shekharbhardwaj/agenticledger)?variant=verified to the badge URL.Vetting this one by hand? Tool Check is an MCP that scores other MCPs. Add it once and ask Claude, ChatGPT, or any MCP client to grade a server, surface CVEs, check the publisher, and suggest safer alternatives — before you install.
https://m8ven.ai/api/mcp/tool-check