Glog

You could spin up Elasticsearch, Logstash, and Kibana. Configure a Loki cluster. Pay Datadog per gigabyte. Set up Fluentd with seventeen plugins.

Or: run one binary. Ship logs. Debug with a live workspace.

GLog is a local-first logging and debugging workspace that runs on SQLite and ships as a single binary. It gives small teams and coding agents a live dashboard, structured log search, rollups, repeated-pattern grouping, exports, and an inbox for notes, tasks, bugs, questions, and handoffs.

30-second demo

# Install
curl -fsSL https://raw.githubusercontent.com/dotcommander/glog/main/scripts/install.sh | sh

# Verify the install and see exact next steps
glog doctor

# Start the server (database creates itself)
glog serve

# In another terminal: register a host, get an API key
glog host register --server http://localhost:6016 --name my-server
# Writes config to ./glog.json — no manual setup

# Ship a log
glog log "Deployment complete" --level info --field version=2.1.0

# Pipe anything
tail -f /var/log/syslog | glog log --stream --level info

Server opens at http://localhost:6016. Dashboard and API, same port.

Install

Binary (recommended):

curl -fsSL https://raw.githubusercontent.com/dotcommander/glog/main/scripts/install.sh | sh
glog doctor

The installer downloads the right macOS/Linux archive, installs glog into /usr/local/bin or ~/.local/bin, warns if that directory is not on PATH, and prints the first server/register/log commands.

You can also grab archives manually from the releases page. macOS, Linux, and Windows builds are published for amd64 and arm64.

Go install:

go install github.com/dotcommander/glog/cmd/glog@latest

From source:

git clone https://github.com/dotcommander/glog && cd glog
make          # Builds frontend + backend, starts on :6016
make install  # Puts glog in ~/go/bin

CLI

glog serve                              # Start server (:6016 by default)
glog serve --addr :8080 --db logs.db    # Custom port and database path
glog serve --durability strict          # SQLite sync profile: strict, balanced, fast

glog host register --name prod-api      # Register host, saves config to ./glog.json
glog host list                          # Show registered hosts

glog log "Something happened"           # Send a log (level: info)
glog log --level error "Disk full"      # Explicit level
glog log --field region=us-east-1 "OK"  # Attach structured fields

journalctl -f | glog log --stream       # Stream stdin, batched at 100 lines
cat app.log | glog log --level warn     # Pipe a file

glog migrate --db glog.db               # Run database migrations
glog inspect db --db glog.db --json     # Offline database diagnostics
glog doctor                             # Check PATH, config, server, and next steps
glog inbox add "Follow up on flaky test" # Leave an agent note/task
glog inbox list                         # Show open agent inbox items
glog inbox done 123                     # Complete an inbox item
glog version                            # Print version and build time

Configuration resolves in order: flags, then GLOG_SERVER/GLOG_API_KEY env vars, then ./glog.json.

For LLM coding agents

Use GLog as your local debugging memory when a task spans commands, browser checks, or multiple agent sessions.

# 1. Find or start the local server.
glog doctor
glog serve

# 2. Register the current workspace once, then reuse ./glog.json.
glog host register --server http://localhost:6016 --name "$(basename "$PWD")"

# 3. Send useful evidence while debugging.
glog log --level info "starting frontend smoke test" \
  --field repo="$(basename "$PWD")" \
  --field command="bun run build"

# 4. Leave durable follow-up work in the agent inbox.
glog inbox add "Recheck SSE reconnect after hub change" \
  --kind task \
  --priority normal \
  --repo "$(basename "$PWD")" \
  --session "$USER-$(date +%Y%m%d-%H%M%S)" \
  --tag sse \
  --body "Evidence belongs in logs. This item is the next action."

# 5. Before finishing, review open work and close completed items.
glog inbox list --status open --repo "$(basename "$PWD")"
glog inbox done 123

Agent rules of thumb:

• Put runtime evidence, command output summaries, errors, and browser observations in logs.
• Put decisions, blocked questions, next tasks, bugs, and handoffs in the inbox.
• Include repo, session, and focused tags so future agents can filter quickly.
• Link the exact error text, fingerprint, or log ID in the inbox body when a task depends on evidence.
• Mark inbox items done when the task is handled; leave blocked questions open with --kind question --priority high.

Full reference: Agent inbox.

API

Hosts

Logs

Agent Inbox

Stream & Export

Auth means Authorization: Bearer glog_v1_<key> — the key you get from host register.

# Create a log via curl
curl -X POST http://localhost:6016/api/v1/logs \
  -H "Authorization: Bearer glog_v1_<key>" \
  -H "Content-Type: application/json" \
  -d '{"level":"error","message":"Payment failed","fields":{"order_id":"abc123"}}'

# Stream live events
curl -N http://localhost:6016/api/v1/events

Under the hood

Go backend. Svelte 5 frontend. SQLite in WAL mode. Everything runs from one binary.

┌─────────────────────────────────────────────────┐
│  CLI (cobra)          Browser (Svelte 5)        │
│  glog log/host/serve  localhost:6016            │
└──────────┬────────────────────┬─────────────────┘
           │                    │
     ┌─────▼────────────────────▼─────┐
     │       HTTP / SSE (chi/v5)      │
     │  handlers → repos (no service  │
     │  layer, no ceremony)           │
     └──────────────┬─────────────────┘
                    │
          ┌─────────▼─────────┐
          │   SQLite (WAL)    │
          │   single file     │
          │   embedded migrations │
          └───────────────────┘

Handlers call repositories directly. The domain layer defines entities and interfaces — no infrastructure imports. Swapping SQLite for PostgreSQL means writing new repository implementations. Nothing else changes.

cmd/glog/           CLI and server entry point
internal/
  domain/           Entities, repository interfaces, pattern matching
  infrastructure/   HTTP handlers, SQLite repos, SSE hub
web/                Svelte 5 frontend (built into web/build/)

Philosophy

Small tools should stay understandable. Your logs belong on your hardware. The database is a single file you can cp to back up. The server is a single binary you can scp to deploy.

GLog is intentionally modest infrastructure: start the binary, point your apps or coding agents at it, inspect the evidence, and leave durable follow-up work in the inbox when the debugging thread needs to continue.

Why not X?

GLog is for teams and coding agents that need operational evidence without a logging platform project. If you need petabyte-scale search across a fleet of thousands, use the big tools. If you want a local, inspectable place to capture logs, spot patterns, and hand off debugging state, this is the one.

Docs

• Go integration — embed glog in your Go services
• Getting started — first install, first server, first log
• Deployment — launchd, Docker, systemd, nginx
• SSE events — real-time event format
• Agent inbox — notes, tasks, questions, and handoffs for coding agents
• Architecture — design decisions
• Invariants — behavior future changes must preserve
• Contributing — commit format, test requirements

License

MIT. Because logs shouldn't require permission.