Telemetry in, evidence out.

Blop is a local-first OpenTelemetry observability tool for teams working with coding agents. One command starts a collector on your machine; point any OTLP exporter at it and query services, traces, logs, findings, and audits from the CLI with stable JSON output. No account, no autonomous actions, no guesswork.

Prompt

Set up Blop (https://docs.blopai.com) on a local checkout of my project. First read the docs and follow Getting Started. Don't change my production CI. Use the project's existing package manager, start Blop's local collector, and connect the project's existing OpenTelemetry traces and logs without adding a Blop-specific SDK. Run the project locally, show me the services, traces, logs, and a Blop audit, then run it again after a small local change so I can see how Blop helps an agent verify its work with exact evidence. I want to understand how easy this is to set up and how it can help coding agents write and audit better code.

Install the CLI Read the docs

OTLP-native · Local SQLite store · Stable --json for agents

Setup

One collector. Three steps to evidence.

Start the monitor, point your OpenTelemetry exporter at it, and ask for evidence. Everything runs on your machine; no account required.

blop traces --service checkout --status error

Step 1 / 3

Start the collector

blop monitor start runs a local OTLP receiver on loopback, backed by SQLite. It prints its endpoints and the database path, so you always know exactly where telemetry lives.

Step 2 / 3

Point your exporter at it

Set OTEL_EXPORTER_OTLP_ENDPOINT and run your app or test suite. Standard OTLP/HTTP traces and logs, JSON or protobuf, from any language. No Blop SDK, no code changes.

Step 3 / 3

Ask for evidence

Query services, traces, logs, and findings with bounded windows and stable JSON. blop audit summarizes the window and cites the identifiers behind every finding.

Plays nicely with

GitHub GitHub Actions Slack Linear CloudEvents OIDC

The evidence loop

Standard telemetry in. Citable evidence out.

Run blop monitor start and point any OpenTelemetry exporter at it: your app, your tests, your CI. Traces and logs land in a local SQLite store, and every query returns a bounded, deterministic answer with stable JSON your coding agent can read.

Any OTLP exporter works; no Blop SDK

Local SQLite store, WAL, and retention

Bounded results, stable JSON, documented exit codes

blop dispatch · checkout-suite

ci.yml · the only change is standard OpenTelemetry config

 env:
   OTEL_EXPORTER_OTLP_ENDPOINT: http://127.0.0.1:27686
   OTEL_SERVICE_NAME: checkout

Findings, not log spelunking

Audits turn telemetry into findings with evidence.

blop audit runs deterministic analysis over the stored window: error traces and their root spans, new error signatures against a baseline, slow traces and spans, log bursts, incomplete traces, and services that stopped emitting. Every finding carries a code, a severity, the trace and span IDs that prove it, and the next command to dig deeper. No model calls, no guessing.

Deterministic; reproducible for the same stored data

Findings cite trace, span, and log identifiers

Gate CI with --fail-on regression

trace 4f1a9c checkout · POST /pay

1 error trace observed Step 1 of 3

finding new_error_signature severity high

Root operation: POST /pay Step 2 of 3

audit --fail-on regression exit 0

Fix verified, no new findings Step 3 of 3

Services and signals

Every service you are changing, in one window.

Services appear as soon as they emit, named by the standard service.name resource attribute. Scan error rates, latency, and log severity per service, then drill from service to trace to correlated logs without leaving the terminal.

Services appear from standard OTLP attributes

Bounded lookbacks and limits by default

Wherever it runs

From laptop to CI, one local collector receives every service you are changing

Integrations

Linear

Wired into the stack you already run

Traces · last 24h 99.7% ok

By service

checkout 99.6% 4.2s p95

billing-api 99.8% 3.8s p95

auth 100.0% 1.4s p95

webhook-worker 99.4% 6.1s p95

services observed

0.0%

traces ok, last 15m

0.0s

median service p95

The loop

Change, rerun, and verify with evidence.

Your coding agent makes the edit, reruns the workload, and queries blop again. Before it declares success, blop audit --fail-on regression compares against the baseline window and exits nonzero on anything new. Proof, not vibes.

Audit · before

3 error traces · 1 latency finding

10 minutes ago

Trace 4f6c9c

root span: payment timeout

6 minutes ago

Audit · after fix

0 error traces · exit 0

2 minutes ago

Findings cite trace IDs, span IDs, timestamps, durations, and status for every claim, so a reviewer can open the exact telemetry behind a conclusion. Blop separates observed facts from derived findings, and never claims causality from correlation alone.

Reproducible before/after comparison

Exit codes your CI and agent can branch on

Walkthrough

Product walkthrough

A short look at the local collector, the queries, and the audit. The recording is not published yet.

Video coming soon

Pricing

Free locally. Pay for the team platform.

The CLI, collector, local store, queries, and audits are free and run entirely on your machine. Hosted billing is early access and not live. Paid bands are a reserved catalog for team history and support as the platform ships; local use never requires an account.

Starter

€199 / mo

Reserved catalog price · billing is not live

For a team keeping one service under continuous observation.

Unlimited local queries & seats
100 reserved service slots
1,000 reserved platform credits / mo
1 reserved project slot
10 reserved scheduled-check slots
Planned integrations: GitHub, Slack

Join early access

Team

Scale

€1,499 / mo

Reserved catalog price · billing is not live

For a portfolio of services with hosted history and audit artefacts.

Unlimited local queries & seats
1,500 reserved service slots
8,000 reserved platform credits / mo
10 reserved project slots
50 reserved scheduled-check slots
Planned integrations: GitHub, Slack, Linear, Jira, Webhooks, Portfolio view

Join early access

Compare all plans

FAQ

The questions teams ask first.

The honest answers about telemetry, agents, data ownership, and what Blop will not do.

Where does my telemetry live?

In a local SQLite database under your platform state directory, with WAL enabled. blop monitor start prints the exact path at startup. Nothing leaves your machine unless you explicitly configure a remote destination.

Do I need a Blop SDK in my app?

No. Blop receives standard OTLP/HTTP traces and logs, JSON or protobuf. Any OpenTelemetry exporter works, in any language. Point it at the local collector and you are done.

Is Blop an AI agent?

No. Blop is a command-line tool with no model loop and no autonomous actions. Your coding agent calls blop like any other command; blop returns deterministic, machine-readable evidence.

What does blop audit actually check?

It performs deterministic analysis over the stored window: error traces and their root spans, new error signatures versus a baseline, slow traces and spans, bursts of warning and error logs, incomplete traces, and services that stopped emitting. No LLM is involved.

How do agents rely on the output?

Every query supports --json with one stable schema (blop.cli/v1): snake_case keys, UTC ISO 8601 timestamps, bounded pages, and documented exit codes. Stdout stays clean; diagnostics go to stderr.

What about the earlier browser test runner?

The intent-based browser runner (blop test, run, watch, list, init, skills) still works during migration and remains documented in a clearly labeled compatibility section. It is no longer the center of the product.

See your first trace in five minutes.

Free and local-first. Install blop, start the monitor, and point your exporter at it. No account needed.

Read the docs Install the CLI