A production-ready AI agent framework built on LangGraph.
Run it in your terminal, embed it in your app, or leave it working on a server — one install, a compiled agent, nothing to wire up.

Most frameworks hand you the parts and wish you luck. Bog Agents hands you a working agent — file tools, a real shell, git, sub-agents, plan mode, and ~90 composable middlewares out of the box — then lets you peel layers away or bolt new ones on as the job demands. Secure-by-default backends, bounded retries, and governed autonomy are the baseline, not an afterthought.

Three packages, one stack:

• bog-agents — the Python SDK. create_agent() returns a compiled LangGraph agent with file tools, a shell, git, sub-agents, plan mode, retry-with-backoff, and ~90 composable middlewares. Governed autonomy is first-class: agent teams (bog_agents.teams), runaway cost caps (bog_agents.cost_ledger), proof-of-work evidence bundles (bog_agents.evidence), evals + guardrails as importable primitives, and an OS-level sandbox (bubblewrap/seatbelt + an egress-allowlist proxy) for LocalShellBackend. Drop-in deepagents compatibility, too — create_deep_agent, DeepAgentState, filesystem permissions, harness/provider profiles.
• bog-agents-cli — a coding agent that lives in your terminal. 125+ slash commands, any LLM, persistent memory, MCP marketplace. Turn the engine up when it matters: /team run (a governed team over a task ledger), /best-of-n (N worktree attempts, rubric-judged winner), /jury (multi-reviewer vote on a diff), /operator (auto difficulty routing), and /effort (real per-provider reasoning knobs). Plus /peat personal scheduler, /qa acceptance-criteria harness, /record + /replay, an in-memory secrets vault, bog-agents drive for scripted runs, and a full headless surface so an AI agent or CI job can drive it without a human at the keyboard. Matte-swamp TUI.
• bog-agents-daemon — the patient watcher. Runs your agents on cron / file-change / webhook / git-push triggers; survives reboots; reports back via Slack / email / GitHub / file / webhook.

Built on LangGraph. MIT-licensed.

---

Why Bog Agents

• Resilient by default. Failures retry with bounded backoff, hung commands time out, and a provider hiccup never kills the run. A crash drops a redacted panic dump so the bug report writes itself.
• Secure where it counts. Secure-by-default backends, an optional OS-level sandbox with a network egress allowlist, a memory-only secrets vault that never touches disk, and tokens written atomically at 0o600 before the rename — no world-readable race window.
• Governed autonomy. Agent teams, best-of-N with a rubric judge, and multi-reviewer diff votes — all bounded by hard cost caps and packaged with proof-of-work evidence, so you can trust a run you didn't watch.
• No ceremony. pipx install bog-agents-cli && bog-agents puts a working agent in front of you in under a minute; pip install bog-agents plus one function call embeds one in your code.
• Composable to the core. ~90 middlewares snap on or off, sub-agents nest, backends swap. The framework gets out of the way as your needs sharpen.

---

Quick install

CLI (most folks start here)

# pipx is recommended — isolated install, clean PATH
pipx install bog-agents-cli

# or with uv (fastest)
uv tool install bog-agents-cli

# or plain pip
pip install bog-agents-cli

Provider extras:

pip install 'bog-agents-cli[anthropic]'        # Claude
pip install 'bog-agents-cli[openai]'           # GPT
pip install 'bog-agents-cli[bedrock]'          # AWS Bedrock
pip install 'bog-agents-cli[all-providers]'    # everything

Set one provider key and ride:

export ANTHROPIC_API_KEY=sk-ant-...     # or OPENAI_API_KEY, GOOGLE_API_KEY, AWS creds, ...
bog-agents --doctor-deep                # one-page health check
bog-agents                              # interactive TUI
bog-agents -p "explain this module" < src/agent.py   # one-shot

No key handy? Point it at a local Ollama model and nothing leaves the machine.

Keep current with /update inside the TUI — it checks PyPI, shows what will download, asks before installing, and upgrades the way you installed (uv tool / pipx / pip).

Daemon (ambient runner)

pip install bog-agents-daemon
bog-agents-daemon run --port 7878

See the daemon README for systemd / Windows-task / launchd installation, the REST API, trigger types, and output targets.

SDK only (for embedding agents in your own app)

pip install bog-agents

from bog_agents import create_agent

agent = create_agent(model="anthropic:claude-sonnet-4-6")
result = await agent.ainvoke({"messages": [{"role": "user", "content": "hi"}]})

See the SDK README for backends, middlewares, deepagents compatibility, and the full provider matrix.

---

What's new in 0.9.x

The 0.9 line is the run from a credible flagship toward 1.0. Each release hardened a different stretch of trail.

• 0.10 — Claude-Code-style auto mode, MCP server, self-verification, and two new SDK primitives. The marquee batch:
  • Permission modes (auto mode). Shift+Tab cycles default → accept-edits → plan; Ctrl+T toggles bypass; a live status indicator; a --permission-mode {default,acceptEdits,plan,bypass,paranoid} flag (plus --dangerously-skip-permissions).
  • bog-agents mcp-server. Expose the agent as an MCP server, so Claude Desktop, Cursor, Zed, or Copilot can delegate a whole coding task to it.
  • /self-review. Fan five reviewer lenses (correctness, security, maintainability, tests, over-claims) over your own diff → SHIP / FIX-FIRST verdict; --fix loops until clean.
  • /ci-fix. Read the branch's CI via gh, ingest the failing-job logs, and diagnose + fix.
  • @codebase semantic search, repo-committed .prompt.md files that auto-register as slash commands, agent-written auto-memories (a remember tool that persists conventions/gotchas to the AGENTS.md cascade), and shell pass-through (!command output now enters the agent's context so it can see what you ran).
  • Governed autonomy — turn the engine up. /team run runs a real agent team over a claimable, dependency-aware task ledger; /best-of-n runs N attempts in isolated git worktrees and keeps the rubric-judged winner; /jury gets a multi-reviewer vote on a diff; /operator lets a cheap judge auto-escalate model + effort (and route hard jobs to butcher / jtbd). All bounded by runaway cost caps so nothing fork-bombs the bill.
  • OS-level sandbox. .bog-agents/sandbox.toml (local_sandbox + network_allowlist) wraps every shell command in bubblewrap (Linux) / seatbelt (macOS), with a hard network cut or a localhost egress-allowlist proxy; require_sandbox fails closed where no launcher exists.
  • Assign-to-bog. The daemon's POST /webhooks/github front door turns an assigned issue / applied label / review comment / red CI run into an agent job (HMAC-verified, fail-closed).
  • SDK: bog_agents.evals (Dataset / scorers / run_evals / assert_pass_rate), bog_agents.guardrails (composable input/output tripwire guardrails + create_agent(guardrails=[...])), bog_agents.teams (ledger + mailbox + run_team), bog_agents.cost_ledger (CostLedger + RunawayCaps), and bog_agents.evidence (proof-of-work bundles), plus the declarative .bog-agents/sandbox.toml and the LocalShellBackend OS sandbox. Hardened by a fresh Senior-Principal-Engineer audit (see REVIEW.md) and a competitive roadmap (ROADMAP.md).
• 0.9.4 — deepagents parity, headless driving, provider resilience. A first-class deepagents compatibility surface — create_deep_agent, DeepAgentState, FilesystemPermission, RubricMiddleware, and HarnessProfile / ProviderProfile keyed by provider:model — so a deepagents user can switch over without rewriting, and back again. A headless command surface (bog-agents command "/help") plus --jsonl structured streaming so agents and CI can drive the CLI non-interactively. Live-tested resilience across Anthropic, AWS Bedrock, and OpenAI.
• 0.9.1 — Bedrock, seamless. Automatic inference-profile resolution, /bedrock fix + /bedrock config, and auto SSO-credential refresh. Point at a model id; the SDK sorts out the rest.
• 0.9.0 — scriptable TUI, compliance, security sweep. bog-agents drive graduated to a full Pilot-backed runner; /compliance auditor with HMAC-sealed reports; a repo-wide security pass.

See CHANGELOG.md for the full history, including the 0.8.x flagship features (/peat, /qa, /record + /replay, the MCP marketplace) carried forward below.

---

The 0.8 flagship features (still here, still sharp)

/peat — your personal assistant

A long-lived in-process sub-agent with a hand-crafted persona. Schedules recurring jobs (cron, @every, @once), runs deep research with a five-phase plan, builds personalized digests from your /qa results and /replay recordings.

/peat schedule "0 9 * * 1-5 | summarize yesterday's QA results"
/peat research "vector databases" --focus pricing,perf
/peat digest --days 7

/qa — adaptive QA harness

Acceptance-criteria-driven QA plans. Ingest from Jira (via your MCP Jira tool), file, JSON, or stdin. Hybrid step model — agent / shell / http / mcp — with verdicts (exit_code, status, contains, regex, json_path). Outputs as Markdown, JSON, stdout, or Jira comment.

/record + /replay — sessions you can edit and re-run

/record captures user prompts, AI responses, and tool calls live. /record stop finalizes to a YAML file with auto-detected variables (Jira IDs, repo URLs, file paths) replaced by ${var} placeholders. /replay run prompts for any unfilled variables and dispatches to the agent.

Vault + Vars

Typed variable system shared by /replay and /qa: string, secret, enum, int, bool. Secrets live only in process memory; nothing persists to disk. Optional read-only OS-keychain bridge.

MCP marketplace

35+ curated servers across 9 categories: github, jira, gitlab, slack, postgres, mongodb, redis, bigquery, snowflake, supabase, aws, azure-devops, terraform, cloudflare, stripe, hubspot, notion, confluence, google-drive, discord, kubernetes, datadog, sentry, and more.

/mcp marketplace          # browse the catalog
/mcp install jira         # install from the catalog
/mcp add my-tool ...      # custom server

---

Repository layout

Path	What
libs/bog-agents/	The Python SDK. Compiled LangGraph agents, 90+ middlewares, pluggable backends, tool bundles, deepagents compatibility.
libs/cli/	The terminal CLI. Textual TUI, 120+ slash commands, MCP marketplace, headless command surface, bog-agents drive scripted runner.
libs/daemon/	The ambient daemon. Cron / file-watch / webhook / git-push triggers, REST API.
libs/acp/	Agent Client Protocol bridge for the Zed editor.
libs/harbor/	Evaluation / benchmark harness (Terminal Bench 2.0).
libs/partners/	Sandbox provider integrations (first-party source today: Daytona).
libs/vscode-extension/	VS Code integration.

Each package has its own pyproject.toml, Makefile, and version. SDK / CLI / daemon are released together on synchronized versions; the rest tag independently.

---

Working from source

git clone https://github.com/bogware/bog-agents
cd bog-agents

# Install + run the CLI from source
cd libs/cli
uv sync --reinstall
uv run bog-agents

Every package answers the same Makefile targets:

make test        # unit tests, no network
make lint        # ruff check + ruff format --diff + ty
make format      # ruff fix + ruff format

CI runs make lint + make test for the SDK, CLI, and daemon on every PR, make test for the satellites (acp, harbor, daytona), and a lockfile-drift check across all packages. The VS Code extension builds on manual dispatch. See .github/workflows/ci.yml.

---

Documentation

Start with docs/ — the full documentation tree:

• Getting Started — install, first run, the five commands you'll use forever
• Cookbook — fifteen task-shaped recipes
• Troubleshooting — every common error, what it means, how to fix it
• Tips & Tricks — power-user moves
• Security Model — what's safe by default, what isn't, threat boundaries
• CLI deep dives: Drive runner · Slash command reference
• SDK guides: Quickstart · Middleware · Tool bundles
• Providers: AWS Bedrock
• Daemon: Quickstart
• Advanced: Expert Rules

Also:

• Per-package READMEs: SDK · CLI · Daemon
• Architecture — CLAUDE.md
• Changelog — CHANGELOG.md
• Issues — https://github.com/bogware/bog-agents/issues

---

License

MIT. See LICENSE.

---

Pass through in harmony.