Agentic Ledger

Runtime observability for AI agents — see exactly what your agent did, why it did it, and what it cost.

Website: agentic-ledger.dev

The numbers are meant to match your provider bill. If they don't, that's a bug we want.

Works with any agent framework, any LLM provider, any model gateway. Zero code changes required. Point your agent at the proxy and everything is captured automatically.

How it works

Agentic Ledger runs as a transparent proxy between your agent and the LLM provider. It intercepts every request and response, assigns it an action_id, stores it, and returns the upstream response unmodified. Your agent never knows the proxy is there. The full picture, with diagrams and a module map for contributors, lives in ARCHITECTURE.md.

Your Agent  →  Agentic Ledger Proxy  →  OpenAI / Anthropic / LiteLLM / any LLM
                      ↓
               SQLite or Postgres
                      ↓
               Live Dashboard + API

Related MCP server: swarm-at

Quick Start

Step 1 — Start the proxy

Two commands, zero config, no terminal held hostage:

uv tool install agentic-ledger    # or: pipx install agentic-ledger, or pip install -U agentic-ledger
agenticledger start     # runs in the background; terminal freed

A tool-managed install (uv tool / pipx) gets its own isolated environment and one unambiguous shim on PATH, so it can never be shadowed by another Python's copy and agenticledger upgrade always means exactly one thing. Plain pip works too; if a machine ever grows competing installs, agenticledger doctor --fix untangles them.

agenticledger start prints the dashboard URL and gives your terminal back — closing the window doesn't stop it. agenticledger status tells you it's up and healthy, agenticledger logs shows what it's doing, agenticledger stop shuts it down. Want a config file anyway? agenticledger init writes a commented one; see Configuration for what goes in it.

Or with Docker (no Python required):

docker run -p 8000:8000 \
  -e AGENTICLEDGER_UPSTREAM_URL=https://api.openai.com \  # optional: omit to route by call format
  -v $(pwd)/data:/data \
  ghcr.io/shekharbhardwaj/agentic-ledger:latest

The image is multi-arch (amd64/arm64), runs as a non-root user, and every release is signed with Sigstore and ships an SBOM. Hardening a shared deployment (TLS, auth keys, redaction, verification)? See the deployment guide.

Using Anthropic / Claude? Nothing to configure: with no upstream set, the proxy routes each call by its wire format, so Anthropic-style calls go to Anthropic and OpenAI-style calls go to OpenAI, side by side through one proxy. Setting an explicit upstream_url (a gateway like LiteLLM or OpenRouter, LM Studio, or a pinned provider) switches to the classic one-proxy-one-provider behavior, mismatch hints included.

Or with docker compose (SQLite by default — see docker-compose.yml):

AGENTICLEDGER_UPSTREAM_URL=https://api.openai.com docker compose up

With uv:

uv add agentic-ledger
AGENTICLEDGER_UPSTREAM_URL=https://api.openai.com uv run python -m agenticledger.proxy

With pip:

python -m venv venv && source venv/bin/activate
pip install -U agentic-ledger
AGENTICLEDGER_UPSTREAM_URL=https://api.openai.com ./venv/bin/python -m agenticledger.proxy

Postgres? Install the extra and set AGENTICLEDGER_DSN:

pip install "agentic-ledger[postgres]"
AGENTICLEDGER_DSN=postgresql://user:password@localhost/agenticledger

Note: the Docker image uses SQLite only. For Postgres with Docker, install via pip instead.

OpenTelemetry? Install the extra and set AGENTICLEDGER_OTEL_ENDPOINT:

pip install "agentic-ledger[otel]"
AGENTICLEDGER_OTEL_ENDPOINT=http://localhost:4318

Proxy starts on http://localhost:8000. Traces are saved to ~/.agenticledger/agenticledger.db when started with agenticledger start (one home for the background service, wherever you launched it from), to agenticledger.db in the current folder when run in the foreground (agenticledger serve / python -m agenticledger.proxy), or to /data/agenticledger.db in Docker.

Step 2 — Point your agent at the proxy

For Claude Code, BMAD, or OpenClaw, one command writes the config for you (backed up, merged, Docker-aware):

agenticledger connect claude-code    # or: bmad, openclaw

For everything else, two changes: set base_url to the proxy and add a session ID header to group calls into a run. Everything else — your API key, model, messages — stays exactly the same.

OpenAI:

from openai import OpenAI

client = OpenAI(
    base_url="http://localhost:8000/v1",  # ← proxy
    api_key="your-openai-key",
    default_headers={"x-agenticledger-session-id": "run-1"},
)

response = client.chat.completions.create(
    model="gpt-4o",
    messages=[{"role": "user", "content": "Research the top 3 AI trends in 2026"}],
)

Anthropic (no upstream config needed: /v1/messages calls route to Anthropic automatically):

import anthropic

client = anthropic.Anthropic(
    base_url="http://localhost:8000",  # ← proxy
    api_key="your-anthropic-key",
    default_headers={"x-agenticledger-session-id": "run-1"},
)

Azure OpenAI: point AzureOpenAI(azure_endpoint="http://localhost:8000") at the ledger with your resource set as the upstream; deployments are priced from the model the response names. See the Azure guide.

AWS Bedrock: install agentic-ledger[bedrock], give the ledger AWS credentials through the standard chain, and point boto3 (endpoint_url) or Claude Code (ANTHROPIC_BEDROCK_BASE_URL) at it; the ledger re-signs each call itself. Both wires are covered: InvokeModel and the modern Converse/ConverseStream APIs. See the Bedrock guide.

LiteLLM / OpenRouter / any gateway:

# Point Agentic Ledger at your gateway
AGENTICLEDGER_UPSTREAM_URL=http://localhost:4000 uv run python -m agenticledger.proxy

# Then point your agent at Agentic Ledger
client = OpenAI(base_url="http://localhost:8000/v1", ...)

Step 3 — Open the dashboard

http://localhost:8000

The web app updates live via WebSocket as calls come in. No refresh needed.

• Loop Lens — every loop run with status (running / flagged / complete / ended / stopped), a cost-per-iteration chart, a block-calls button that refuses a running loop's further calls at the wall (and an allow-calls-again to lift it; the agent being blocked cannot), per-iteration breakdowns, and plain-English explanations of every flag. Pick any two runs with ⇆ to diff them side by side — cost, iterations, calls, flags, duration with signed deltas, plus a prompt drift diff showing exactly what changed in the system prompt and opening instruction between the runs.

• Sessions — every session with three views: expandable call cards (response, thinking, tool calls, cache tokens, interaction badges), a Flow DAG of agent handoffs, and a Trace waterfall with real parent links from the loop engine. Session cards say whose they are and how they're doing at a glance: team badge, red "N failed" for real errors, amber "N blocked" for budget walls, purple tiles for replays. Hover a card for one-click delete.

• Replay the whole run — the question that decides a model switch isn't "how did it handle one call?" but "would my loop have survived?" Pick a run or session, pick a destination (a local model is free), and every step re-runs with its original inputs. You get a report card, not homework: "34 / 40 moments matched", the fumbles named ("dropped the tools"), and the cost both ways. Each step is a real captured moment replayed honestly — after step one a different model would have steered a different conversation, so the ledger compares moments, not fairy tales.

• In your pocket — agenticledger share opens an https tunnel you own (via cloudflared, no account), prints the pairing link, and draws a QR in the terminal: point your phone's camera and the dashboard is in your hand, kill switch and ceilings included. --wifi for a same-network link, --rotate to un-pair every device, or press Pair a device in the dashboard's ⚿ panel. Local machines never need a key; everyone else meets the auto-generated pairing key. The dashboard fits a phone: one pane at a time, a back button, prev/next arrows to flip between runs.

• Named instances — agenticledger start --name demo --port 8003 runs a second ledger beside your everyday one: own state, own database, its dashboard wears an amber name chip so it can never pass for the real thing. stop, status, logs, share, and run all take --name.

• The spend meter — a run's detail reads its money live: spent so far, burning $/h, "at this pace $Y by 8:00 AM". Give any run a cost ceiling and the proxy refuses further calls the moment spend reaches it (amber, costing nothing) until you raise or clear it; the ceiling survives restarts and guards auto-detected loops too. A webhook alert fires at 80%.

• Names, pins, projects — call it "the overnight auth fix" instead of cc-73a26366, ★ pin what matters to the top, file work under a project and the Sessions view reads as sections: a heading per project, its sessions beneath, the unfiled pile last. A run filed under a project files its sessions with it.

• Settings — the ⚙ shows what the proxy is actually running with: config file in effect, upstream, budgets, replay targets, each row labeled file / env / default. Read-only, secrets hidden.

• Replay & what-if — open any call and ↻ Replay it: pick a destination (the panel lists what your local server actually has loaded), and the exact captured prompt re-executes there — same provider, the other one, or a free local model via LM Studio; tool calls, schemas, and system prompts are translated between the Anthropic and OpenAI wire formats automatically. Works even on calls your own budget blocked — the wall can say no and you can still see what would have happened, for $0. Replays tie back to their original with ↩ Open original. The what-if box answers the cheaper question first: reprice any run or session on another model with pure math, no API calls. (Configure AGENTICLEDGER_REPLAY_API_KEY and/or the per-provider AGENTICLEDGER_REPLAY_*_KEY targets.)

• Reports — where the money goes: spend per day, model mix with latency p50/p95/p99, per-agent totals, a by-team table with each team's spend against its card's daily allowance ("who ran dry?" in one glance), and cache savings — what your prompt-cache traffic would have cost at full input rates versus what it actually cost. Errors and blocks are counted apart everywhere: red = something broke, amber = the ledger refused on purpose — a healthy wall never makes a healthy agent look sick

• Search — full-text search across all sessions by prompt, output, agent name, or user ID

Configuration

agenticledger init writes agenticledger.toml with every option commented. Uncomment what you need — a working setup looks like this:

[proxy]
port = 8000
upstream_url = "https://api.anthropic.com"
db = "sqlite:///agenticledger.db"

[keys]
# Prefer *_file: the file's contents are the key, so no secret lives in
# this file or your shell history (chmod 600 the key file).
api_key_file = "~/.agenticledger/api.key"       # dashboard/admin access
ingest_key_file = "~/.agenticledger/ingest.key" # closes the open relay

[budgets]
daily = 25.0          # whole-ledger daily ceiling, USD
session = 5.0         # per-session ceiling

[replay]
# Free local replay via LM Studio (any key works there):
openai_url = "http://localhost:1234"
openai_key = "lm-studio"

Three rules:

1. The file is found in this order: AGENTICLEDGER_CONFIG, then ./agenticledger.toml (the folder you start from), then ~/.agenticledger/config.toml. First match wins; the startup banner names the file in effect.

2. Anything typed in the command beats the file. Env vars override per-setting (AGENTICLEDGER_PORT=9000 agenticledger start uses 9000 for that run without touching the file) — which is also why Docker and CI setups configured by env vars are unaffected.

3. Changes apply on restart (agenticledger stop then start).

Every setting in the environment-variable reference below has a config-file home; an [env] section passes any other AGENTICLEDGER_* variable through verbatim.

Providers, step by step

Every provider below rides the same proxy; the only thing that changes is which base URL you point at it. Each recipe assumes the proxy is up (agenticledger start) and ends with the same check: run one call, open http://localhost:8000, and see it in Sessions.

OpenAI (and any OpenAI-compatible API)

1. Point the client at the proxy:

   export OPENAI_BASE_URL=http://localhost:8000/v1

2. Keep your OPENAI_API_KEY exactly as it was — the proxy passes your auth header through untouched.

3. Make a call; it appears in Sessions with an O mark.

Anthropic

1. Point the client at the proxy:

   export ANTHROPIC_BASE_URL=http://localhost:8000

2. Keep your ANTHROPIC_API_KEY as it was.

3. Make a call; it appears with an A mark. No upstream config needed — the proxy routes Anthropic-shaped calls to Anthropic by wire format.

AWS Bedrock (direct capture)

1. Give the ledger AWS credentials of its own through the standard chain (env vars, ~/.aws profile, or an instance role) scoped to bedrock:InvokeModel and bedrock:InvokeModelWithResponseStream, then install the extra and restart:

   pip install "agentic-ledger[bedrock]"
   agenticledger stop && agenticledger start

2. Check the ⚙ Settings panel: the Bedrock row should read "signing as the ledger in ".

3. Point the client at the proxy — Claude Code:

   export CLAUDE_CODE_USE_BEDROCK=1
   export ANTHROPIC_BEDROCK_BASE_URL=http://localhost:8000

   boto3: boto3.client("bedrock-runtime", endpoint_url="http://localhost:8000").

4. Make a call; it appears with an orange B mark. The ledger strips the caller's identity and re-signs with its own credentials. Full guide: docs/integrations/bedrock.md.

Azure OpenAI

1. Set the upstream to your resource:

   agenticledger config set proxy.upstream_url https://<resource>.openai.azure.com
   agenticledger stop && agenticledger start

2. Point the client's Azure endpoint at http://localhost:8000; keep your api-key header as it was.

3. Calls are tagged azure-openai and priced by the model the RESPONSE names, so deployment aliases can't hide the real model. Full guide: docs/integrations/azure-openai.md.

Local models (LM Studio, Ollama with the OpenAI API)

1. Set the upstream to the local server:

   agenticledger config set proxy.upstream_url http://localhost:1234
   agenticledger stop && agenticledger start

2. export OPENAI_BASE_URL=http://localhost:8000/v1 in the agent.

3. Calls appear with a purple mark and $0 cost. Full guide: docs/integrations/lm-studio.md.

Gateways (OpenRouter, LiteLLM)

1. Set the upstream to the gateway:

   agenticledger config set proxy.upstream_url https://openrouter.ai/api
   agenticledger stop && agenticledger start

2. export OPENAI_BASE_URL=http://localhost:8000/v1; keep the gateway key as it was.

3. Gateway-prefixed model ids ("anthropic/claude-...") price correctly via substring matching. Guides: openrouter.md, litellm.md.

Framework-specific recipes (CrewAI, LangGraph, AutoGen, Vercel AI SDK, pydantic-ai, and more) live in docs/integrations/.

Coding agents — Claude Code, Ralph loops & friends

Claude Code (and most coding agents) can be pointed at the proxy with a single environment variable — no headers, no code changes:

agenticledger start

export ANTHROPIC_BASE_URL=http://localhost:8000
claude

No upstream config needed: calls route to the provider matching their wire format.

Agentic Ledger fingerprints Claude Code traffic automatically: every call is tagged framework=claude-code, and instead of one undifferentiated bucket, each Claude Code session appears under its real session UUID (the same id claude --resume shows), with prompt-cache reads/writes captured and priced correctly — cache traffic is where most of a coding agent's real spend lives.

Want a loop filed under a name you chose? Put one word in front of the command you already run:

agenticledger run nightly-digest -- python agent.py

Your command runs exactly as before; its LLM calls land on the run tile named nightly-digest, and each launch counts as the next iteration, so tomorrow's run joins the same tile. Nothing in your agent's code changes. Add --project acme to file the run under a dashboard project as it starts.

Running an overnight loop (Ralph-style while :; do cat PROMPT.md | claude -p; done)? The same command with loop flags re-executes your command each iteration, attributes every call to the run (via the base URL, no headers needed), and stops on a completion promise, a budget ceiling, or the iteration cap:

AGENTICLEDGER_UPSTREAM_URL=https://api.anthropic.com \
AGENTICLEDGER_COMPLETION_PROMISE="ALL TASKS COMPLETE" \
uv run python -m agenticledger.proxy

agenticledger run overnight --max-iterations 50 --budget 25 -- \
  claude -p "$(cat PROMPT.md)" --dangerously-skip-permissions

Each iteration shows up as iteration N of the run in /api/runs; when the agent prints the completion promise in a response, run status flips to complete and the loop exits with a cost/token summary. The word after run is the run's name; without one the run is named after the folder and the minute (myproject-0819-1936). Rerunning the same name continues its iteration count instead of restarting at 1. Any existing loop script works too — poll GET /api/runs/{run_id} yourself, or let the proxy's budgets (AGENTICLEDGER_BUDGET_DAILY=25.00) hard-stop a runaway loop.

Iterating on the prompt? Rerun and use ⇆ compare in the Loop Lens to diff the two runs — cost, iterations, calls, and flags side by side — so "did the new prompt actually help" gets a number instead of a feeling.

The same recipe works for any client with a base-URL override (Codex CLI, opencode, OpenClaw, LiteLLM-based stacks) — set the OpenAI/Anthropic base URL to the proxy and traffic is captured; add x-agenticledger-* headers when you want explicit attribution.

OTel-native tools (Gemini CLI, Codex [otel], AutoGen/AG2, Pydantic AI, Vercel AI SDK) don't need the proxy at all — point their OTLP exporter at the ledger and GenAI spans are ingested directly:

export OTEL_EXPORTER_OTLP_ENDPOINT=http://localhost:8000

Both OTLP/HTTP encodings are accepted: JSON always, protobuf when the [otel] extra is installed (the Docker image includes it). gRPC exporters should switch to HTTP: OTEL_EXPORTER_OTLP_PROTOCOL=http/protobuf.

Framework guides — one per integration in docs/integrations: Claude Code, Codex CLI, opencode, OpenClaw, BMAD-METHOD, LangGraph/LangChain, CrewAI, OpenAI Agents SDK, Gemini CLI, AutoGen/AG2, Pydantic AI, Vercel AI SDK, LiteLLM, OpenRouter, and LM Studio (fully offline: local model, local ledger).

Production deployment — TLS termination, auth keys, redaction, image signature/SBOM verification, enterprise mirrors, and scaling guidance in docs/deployment.md.

The numbers

Measured, not promised. Reproduce them with python scripts/loadtest.py --calls 2000 --seed 1000000 (Apple M-series MacBook, SQLite backend; re-measured on 0.10 with the provider adapter architecture in place — same numbers, 2,580 to 2,800 calls/s on both):

The honest caveats: the session list aggregates every session on every load, so it grows with total history; the report window uses a timestamp index, so it grows with the window's traffic, not the table. Your agent's provider latency (hundreds of ms per call) dwarfs the proxy's overhead by an order of magnitude. Postgres numbers vary with your server; the same script measures them with --dsn. Cost math has its own guardrails and a five-minute parity check against your provider console: see docs/accuracy.md.

What gets captured

Every LLM call is stored with:

API reference

Examples:

# All calls in a session
curl http://localhost:8000/session/run-1

# Search across all sessions
curl "http://localhost:8000/api/search?q=failed+to+connect"

# Download JSON audit trail (includes an integrity tag; keyed HMAC when configured)
curl http://localhost:8000/export/run-1 -o audit-run-1.json

# Printable HTML report — open in browser, print to PDF
open http://localhost:8000/export/run-1/report

MCP server

Agentic Ledger exposes its captured data as an MCP (Model Context Protocol) tool server at POST /mcp. Point Claude Desktop, Cursor, or any MCP-compatible client at it to query traces directly from your AI assistant.

Tools available:

Configure in claude_desktop_config.json (HTTP, against a running proxy):

{
  "mcpServers": {
    "agenticledger": {
      "url": "http://localhost:8000/mcp"
    }
  }
}

Or as a stdio subprocess — for clients that launch servers as commands (no running proxy required; reads the same database):

{
  "mcpServers": {
    "agenticledger": {
      "command": "agenticledger",
      "args": ["mcp"],
      "env": { "AGENTICLEDGER_DSN": "sqlite:////absolute/path/to/agenticledger.db" }
    }
  }
}

If AGENTICLEDGER_API_KEY is set, pass it as a header:

{
  "mcpServers": {
    "agenticledger": {
      "url": "http://localhost:8000/mcp",
      "headers": { "x-agenticledger-api-key": "your-key" }
    }
  }
}

Once connected, you can ask your assistant things like:

• "What did the SearchAgent do in the last session?"

• "Show me all calls that mentioned rate limit errors"

• "What was the total cost of session run-abc123?"

Configuration reference

Every variable below can also live in agenticledger.toml — see Configuration for the file, the search order, and the env-always-wins rule.

Environment variables

Core:

Cost budgets — block calls that exceed a spend limit (returns HTTP 429):

Rate limits — block calls that exceed request frequency (returns HTTP 429, sliding 60-second window):

Loop engine — every call is stitched into ReAct threads (thread_id, step_index, prev_action_id) and fresh-context loop iterations are grouped into runs, with stuck-loop detection:

Alerts — POST to your webhook when a threshold is breached (does not block calls — see Alerts):

OpenTelemetry — emit spans to any OTLP-compatible collector (requires pip install "agentic-ledger[otel]" — see OpenTelemetry export):

Pricing overrides — override or extend the built-in per-token pricing table (merged at startup):

Common startup examples

# Local dev — OpenAI (default)
AGENTICLEDGER_UPSTREAM_URL=https://api.openai.com uv run python -m agenticledger.proxy

# Local dev — Anthropic
AGENTICLEDGER_UPSTREAM_URL=https://api.anthropic.com uv run python -m agenticledger.proxy

# Local dev — LiteLLM gateway (any model)
AGENTICLEDGER_UPSTREAM_URL=http://localhost:4000 uv run python -m agenticledger.proxy

# Production — Postgres + auth + budgets + rate limits + alerts
AGENTICLEDGER_UPSTREAM_URL=https://api.openai.com \
AGENTICLEDGER_DSN=postgresql://user:password@localhost/agenticledger \
AGENTICLEDGER_API_KEY=my-secret \
AGENTICLEDGER_BUDGET_DAILY=20.00 \
AGENTICLEDGER_BUDGET_SESSION=2.00 \
AGENTICLEDGER_RATE_LIMIT_SESSION_RPM=20 \
AGENTICLEDGER_RATE_LIMIT_USER_RPM=60 \
AGENTICLEDGER_ALERT_WEBHOOK_URL=https://hooks.slack.com/services/xxx/yyy/zzz \
AGENTICLEDGER_ALERT_COST_PER_CALL=0.50 \
AGENTICLEDGER_ALERT_DAILY_SPEND=15.00 \
uv run python -m agenticledger.proxy

When AGENTICLEDGER_API_KEY is set, pass it to access protected endpoints:

# Header
curl -H "x-agenticledger-api-key: my-secret" http://localhost:8000/session/run-1

# Query param (browser)
http://localhost:8000?api_key=my-secret

Scoped API tokens (RBAC)

The master key is convenient but coarse. For team access, mint scoped, revocable tokens with roles instead of sharing the master secret. Tokens are random secrets shown once at creation; only their SHA-256 hash is stored.

Roles are hierarchical:

# Mint a viewer token (admin only — use the master key to bootstrap)
curl -X POST http://localhost:8000/api/tokens \
  -H "x-agenticledger-api-key: my-secret" \
  -H "content-type: application/json" \
  -d '{"name": "grafana-readonly", "role": "viewer", "expires_in_days": 90}'
# → {"token_id": "...", "token": "agl_…", "role": "viewer", ...}  (token shown once)

# Use it (Bearer header, x-agenticledger-token, or ?token=)
curl -H "Authorization: Bearer agl_…" http://localhost:8000/api/sessions

# List and revoke
curl -H "x-agenticledger-api-key: my-secret" http://localhost:8000/api/tokens
curl -X DELETE -H "x-agenticledger-api-key: my-secret" http://localhost:8000/api/tokens/<token_id>

Auth is enforced only when AGENTICLEDGER_API_KEY is set; the master key is the admin bootstrap for minting tokens. The live /ws feed accepts the same credentials (?api_key=, ?token=, Authorization: Bearer, or x-agenticledger-token) and rejects unauthenticated connects with close code 1008 — the dashboard forwards its page credential to the socket automatically.

Request headers

Pass these from your agent on each LLM call. All optional. They enrich captured data, power the Flow tab, and enable per-dimension budgets and rate limits.

Single agent — fully annotated:

from openai import OpenAI

client = OpenAI(
    base_url="http://localhost:8000/v1",
    api_key="your-openai-key",
    default_headers={
        "x-agenticledger-session-id":  "run-abc123",
        "x-agenticledger-user-id":     "user-42",
        "x-agenticledger-agent-name":  "researcher",
        "x-agenticledger-app-id":      "my-app",
        "x-agenticledger-environment": "production",
    },
)

Multi-agent system — tracking handoffs:

from openai import OpenAI

# Orchestrator
orchestrator_client = OpenAI(
    base_url="http://localhost:8000/v1",
    api_key="your-openai-key",
    default_headers={
        "x-agenticledger-session-id":  "run-abc123",
        "x-agenticledger-agent-name":  "orchestrator",
    },
)

# Researcher (receives handoff from orchestrator)
researcher_client = OpenAI(
    base_url="http://localhost:8000/v1",
    api_key="your-openai-key",
    default_headers={
        "x-agenticledger-session-id":   "run-abc123",
        "x-agenticledger-agent-name":   "researcher",
        "x-agenticledger-handoff-from": "orchestrator",
        "x-agenticledger-handoff-to":   "researcher",
    },
)

The Flow tab renders orchestrator → researcher as a DAG with cost and latency on each node.

OpenAI Agents SDK (openai-agents) — per-agent clients:

The openai-agents SDK uses its own internal OpenAI client. To pass Agentic Ledger headers you need to create a client per agent using OpenAIResponsesModel and set it as the agent's model.

import uuid
import os
from openai import AsyncOpenAI
from agents import Agent
from agents.models.openai_responses import OpenAIResponsesModel

SESSION_ID = f"run-{uuid.uuid4().hex[:8]}"  # one per execution
BASE_URL = os.getenv("OPENAI_BASE_URL")      # e.g. http://localhost:8000/v1

def al_model(agent_name: str, model: str = "gpt-4o-mini",
             handoff_from: str | None = None, handoff_to: str | None = None):
    """Create a model instance that sends Agentic Ledger metadata headers."""
    if not BASE_URL:
        return model  # proxy not configured — use default client
    headers = {
        "x-agenticledger-session-id": SESSION_ID,
        "x-agenticledger-agent-name": agent_name,
    }
    if handoff_from:
        headers["x-agenticledger-handoff-from"] = handoff_from
    if handoff_to:
        headers["x-agenticledger-handoff-to"] = handoff_to
    client = AsyncOpenAI(base_url=BASE_URL, api_key=os.getenv("OPENAI_API_KEY", ""),
                         default_headers=headers)
    return OpenAIResponsesModel(model=model, openai_client=client)

planner = Agent(name="PlannerAgent", model=al_model("PlannerAgent", handoff_to="SearchAgent"), ...)
searcher = Agent(name="SearchAgent",  model=al_model("SearchAgent",  handoff_from="PlannerAgent", handoff_to="WriterAgent"), ...)
writer   = Agent(name="WriterAgent",  model=al_model("WriterAgent",  handoff_from="SearchAgent",  handoff_to="EmailAgent"), ...)
emailer  = Agent(name="EmailAgent",   model=al_model("EmailAgent",   handoff_from="WriterAgent"), ...)

Each agent's calls are tagged with its name and pipeline position. The Flow tab renders the full PlannerAgent → SearchAgent → WriterAgent → EmailAgent DAG automatically.

Why per-agent clients? set_default_openai_client() sets a single global client — fine for single-agent apps, but it can't carry different agent_name or handoff_* headers per agent in a multi-agent system. Per-agent OpenAIResponsesModel instances are the correct approach.

Alerts

Agentic Ledger fires a POST to your webhook URL when a threshold is breached. You connect it to whatever you already use — Slack, PagerDuty, Discord, email, or a custom endpoint. Agentic Ledger sends the payload; the integration is on your side.

Payload format:

{
  "type":       "high_cost",
  "message":    "Single call cost $0.1842 exceeded threshold $0.10",
  "value":      0.1842,
  "threshold":  0.10,
  "action_id":  "a1b2c3d4-...",
  "session_id": "run-1",
  "agent_name": "researcher",
  "timestamp":  "2026-04-03T12:00:00+00:00"
}

Alert types:

Team cards — one proxy, many teams

Think allowance cards: you keep the one real provider key, and hand each team a card of its own. Each card opens the proxy, stamps every call with the team's name, and can carry its own daily budget — when marketing hits $10, only marketing gets blocked (with an honest Retry-After).

curl -X POST http://localhost:8000/api/tokens \
  -H "x-agenticledger-api-key: $ADMIN_KEY" -H 'content-type: application/json' \
  -d '{"name": "marketing", "role": "ingest", "budget_daily": 10.00}'

The response shows the card once — the ledger stores only its hash. The team puts it in x-agenticledger-ingest-key instead of the shared key; Reports gains a by-team table with errors, blocks, and spend-today against each card's allowance. Revoke a card with DELETE /api/tokens/{token_id} and only that team is affected — from that instant the card gets a final 403 ("the answer is no"), which agents accept without retry storms. Paste a card into the dashboard's ⚿ panel by mistake and it tells you, in plain words, that cards open the relay, not the dashboard.

Budgets vs alerts:

• Budgets (AGENTICLEDGER_BUDGET_*) — block the call before it reaches the LLM. Agent gets HTTP 429.

• Alerts (AGENTICLEDGER_ALERT_*) — the call goes through, you get notified after.

Slack — create an Incoming Webhook and point AGENTICLEDGER_ALERT_WEBHOOK_URL at it. Add AGENTICLEDGER_DIGEST_HOUR=8 and the same webhook also gets a daily good-morning digest: last-24h spend, cache savings, and the top models and agents.

PagerDuty — use the Events API v2 URL or a thin adapter that maps type → PagerDuty severity.

Discord — use a Discord channel webhook URL directly.

Custom — any HTTP endpoint that accepts a JSON POST.

OpenTelemetry export

Agentic Ledger can emit every intercepted LLM call as an OTel span to any OTLP-compatible collector: Grafana Tempo, Jaeger, Honeycomb, Datadog, Dynatrace, or any vendor that supports OTLP/HTTP.

Install the extra (Docker image includes OTel — no extra step needed when using Docker):

pip install "agentic-ledger[otel]"
# or
uv add "agentic-ledger[otel]"

Configure:

Example — Grafana Tempo:

AGENTICLEDGER_UPSTREAM_URL=https://api.openai.com \
AGENTICLEDGER_OTEL_ENDPOINT=http://localhost:4318 \
AGENTICLEDGER_OTEL_SERVICE_NAME=my-agent \
uv run python -m agenticledger.proxy

Example — Honeycomb:

AGENTICLEDGER_OTEL_ENDPOINT=https://api.honeycomb.io \
AGENTICLEDGER_OTEL_HEADERS=x-honeycomb-team=YOUR_API_KEY,x-honeycomb-dataset=llm-traces \
uv run python -m agenticledger.proxy

Span attributes emitted (GenAI semantic conventions):

Spans are grouped into traces by session_id — all calls in a session appear as one trace in your backend. Parent-child relationships follow x-agenticledger-parent-action-id. Error spans (status_code != 200) are marked with StatusCode.ERROR.

Compliance export

Every session can be exported as an integrity-tagged audit trail — useful for regulated industries, internal audits, or passing traces to external tools.

# Machine-readable JSON with an integrity tag over the calls array
curl http://localhost:8000/export/run-1 -o audit-run-1.json

# Printable HTML — open in browser and print to PDF
open http://localhost:8000/export/run-1/report

The JSON export carries an integrity tag over the calls array. By default this is a sha256 checksum — it catches accidental corruption but is not a signature (anyone who edits the calls can recompute it). Set AGENTICLEDGER_EXPORT_HMAC_KEY to switch to a keyed hmac-sha256 tag, which is tamper-evident: a recipient holding the key can detect any modification, and the tag cannot be forged without the key.

Releasing

Tagging a version triggers the full release pipeline automatically:

git tag v0.2.0
git push origin v0.2.0

This runs three jobs:

1. Docker — builds ghcr.io/shekharbhardwaj/agentic-ledger:{version} and :latest for linux/amd64 + linux/arm64, pushes with SBOM + provenance attestations, signs the digest with Sigstore cosign (keyless), and mirrors to Docker Hub when the DOCKERHUB_USERNAME/DOCKERHUB_TOKEN secrets are configured

2. PyPI — builds and publishes agentic-ledger=={version} to PyPI using trusted publishing (no API token needed), with PEP 740 attestations

3. GitHub Release — creates a release with auto-generated changelog and attaches the image SBOM (SPDX)

First-time PyPI setup (one time only):

1. Go to pypi.org/manage/account/publishing

2. Add a new pending publisher:

   PyPI project name:  agentic-ledger
   Owner:              ShekharBhardwaj
   Repository:         AgenticLedger
   Workflow name:      release.yml
   Environment name:   pypi

3. Create a pypi environment in GitHub: repo → Settings → Environments → New environment → name it pypi

4. That's it — no secrets needed

Troubleshooting

Start here: agenticledger doctor. One command prints the whole truth Add --fix and it applies the fixes it names: shadow installs evicted with their own interpreter's pip, the PATH prepend offered for cleanup, then a second diagnostic pass. of your machine: every install on PATH and who shadows whom, which Python owns each one and whether it can actually run (wrong-architecture wheels and missing dependencies caught by a real import probe), what the background service is serving, and a fix-it command per finding. Most of the problems below diagnose themselves with it.

Old version / commands or env vars named agentledger (no "ic") — you're running a pre-0.4 release, most likely from a venv that already had the package installed: plain pip install agentic-ledger says "requirement already satisfied" and does NOT upgrade. Run agenticledger upgrade — it uses the Python environment that owns the install, so there's no guessing which pip is the right one — then restart. The proxy prints its version on the first line at startup, and curl localhost:8000/health reports it too. Since 0.4.0 everything is named agenticledger — see the migration notes in the CHANGELOG.

Replay fails with 401 "invalid x-api-key" — AGENTICLEDGER_REPLAY_API_KEY needs a real provider API key from console.anthropic.com (or platform.openai.com). A Claude Code subscription login is not an API key and cannot be used. No key? Replay for free against a local model — see the LM Studio guide.

incompatible architecture (have 'arm64', need 'x86_64') on macOS — your terminal is running under Rosetta, so Python picks its x86_64 slice while pip installed arm64 native wheels. Check with arch (should print arm64 on Apple Silicon). Quick fix: prefix the command with arch -arm64. Permanent fix: uncheck "Open using Rosetta" on your terminal app, use an Apple Silicon build of your editor, and restart any long-lived tmux server.

module 'httpx' has no attribute 'AsyncClient' — fixed in 0.3.0-alpha.2; upgrade with pip install --upgrade agentic-ledger.

Port 8000 already in use — another proxy instance (or app) is running; stop it or set AGENTICLEDGER_PORT.

401 OAuth access token has expired from Claude Code — the proxy passed Anthropic's answer through unmodified; re-authenticate with claude → /login. Errored calls are still captured, so you'll see the 401 in the dashboard.

/ answers 404 "Web app not built" — you're running from a source checkout without the web-app build. cd dashboard-app && npm ci && npm run build and restart. PyPI and Docker installs always include the app.

License

MIT

mcp-name: io.github.ShekharBhardwaj/agentic-ledger