pi-think — the pi agent with custom think enabled

pi-think is a fork of the pi coding agent that adds one feature on top of an otherwise untouched upstream: a startup-selected think-tool reasoning mode. With it, the model's chain-of-thought is externalized into a plaintext scratchpad tool call, so the full reasoning becomes visible — live in the TUI, and persisted in the session file — even for models whose native thinking is hidden, summarized, or encrypted. Everything below the separator is upstream pi documentation.

What it does

Frontier-model APIs increasingly conceal chain-of-thought: summarized thinking, or fully encrypted thinking blocks (opaque signature, empty text). pi-think operationalizes the scratchpad-tool injection method inside a full coding agent:

• A benign free-text scratchpad tool (default name think, parameter thoughts) is registered. Tool arguments are plaintext-visible to the API caller by protocol necessity — function calling cannot work otherwise.
• The model is steered (system prompt + a forced first call per user prompt) to write its step-by-step reasoning into thoughts, and the TUI renders it like native thinking. The CoT persists in the session transcript as ordinary tool calls, so it stays readable across /model switches and is saved in the session file.

The mechanism is dialect-generic: request shaping keys off the model's API family (anthropic-messages / openai-completions / openai-responses), never a specific provider, so it works with any tool-calling model on those dialects — including Anthropic-compatible gateways and OpenAI-compatible endpoints configured in ~/.pi/agent/models.json. No provider is hardcoded.

Install

git clone https://github.com/ict-agent/pi-think.git
cd pi-think
npm install
npm run build

# run from the repo:
node packages/coding-agent/dist/cli.js --reasoning-mode think-tool

# or put the `pi` binary on your PATH:
(cd packages/coding-agent && npm link)
pi --reasoning-mode think-tool

Requires Node.js (see upstream pi's prerequisites). All stock pi configuration (~/.pi/agent, providers, models, themes, extensions) works unchanged — pi-think only adds the reasoning-mode flags.

Usage

pi --reasoning-mode think-tool --model <provider>/<model>   # CoT visible, live
pi --reasoning-mode think-tool --think-tool-name deep_think # custom scratchpad name
pi --reasoning-mode native                                  # stock pi (default)
PI_WIRE_LOG=/tmp/wire.jsonl pi --reasoning-mode think-tool  # + raw request/response log

• --reasoning-mode is fixed at startup; there is no runtime switching. native is unmodified pi.
• In think-tool mode, Anthropic-dialect reasoning models get thinking: {"type": "disabled"} pinned on the wire, and the scratchpad call is forced via tool_choice on fresh user prompts (continuation turns run auto, so thinking interleaves with real tool use). The same forcing applies on the OpenAI Chat and Responses dialects.
• A scratchpad call truncated by the output token limit is recorded with its salvaged partial arguments and the model is asked to continue — reasoning length is unbounded.
• If a model writes its visible reply into the scratchpad and ends the turn with an empty message (observed with haiku-class models), the harness enqueues a one-shot nudge telling it to write the actual reply.
• PI_WIRE_LOG appends every raw LLM request/response body (JSONL; SSE responses reassembled) — the on-the-wire ground truth for exactly what thinking/tool payloads were sent and received.

Change map vs upstream

• packages/agent: AgentTool.salvageTruncatedArgs + truncated-message salvage path in the agent loop.
• packages/ai: anthropic compat flag supportsPromptCaching (providers whose non-Claude upstreams reject cache_control).
• packages/coding-agent: core/tools/think.ts (scratchpad tool + TUI renderer), core/think-tool-mode.ts (dialect-generic payload shaping), core/wire-log.ts (PI_WIRE_LOG), empty-answer nudge in agent-session.ts, --reasoning-mode / --think-tool-name CLI flags, sdk wiring.

---

New issues and PRs from new contributors are auto-closed by default. Maintainers review auto-closed issues daily. See CONTRIBUTING.md.

Pi Agent Harness

This is the home of the Pi agent harness project including our self extensible coding agent.

• @earendil-works/pi-coding-agent: Interactive coding agent CLI
• @earendil-works/pi-agent-core: Agent runtime with tool calling and state management
• @earendil-works/pi-ai: Unified multi-provider LLM API (OpenAI, Anthropic, Google, …)

To learn more about Pi:

• Visit pi.dev, the project website with demos
• Read the documentation, but you can also ask the agent to explain itself

All Packages

Package	Description
@earendil-works/pi-telemetry	Vendor-neutral telemetry contracts, reference adapter, conformance tests, and typed schemas
@earendil-works/pi-ai	Unified multi-provider LLM API (OpenAI, Anthropic, Google, etc.)
@earendil-works/pi-agent-core	Agent runtime with tool calling and state management
@earendil-works/pi-coding-agent	Interactive coding agent CLI
@earendil-works/pi-tui	Terminal UI library with differential rendering

For Slack/chat automation and workflows see earendil-works/pi-chat.

Permissions & Containerization

Pi does not include a built-in permission system for restricting filesystem, process, network, or credential access. By default, it runs with the permissions of the user and process that launched it.

If you need stronger boundaries, containerize or sandbox Pi. See packages/coding-agent/docs/containerization.md for three patterns:

• Gondolin extension: keep pi and provider auth on the host while routing built-in tools and ! commands into a local Linux micro-VM.
• Plain Docker: run the whole pi process in a local container for simple isolation.
• OpenShell: run the whole pi process in a policy-controlled sandbox.

Contributing

See CONTRIBUTING.md for contribution guidelines and AGENTS.md for project-specific rules (for both humans and agents). Longer term plans for Pi can also be found in RFCs.

Development

npm install --ignore-scripts  # Install all dependencies without running lifecycle scripts
npm run build         # Refresh model data, then build all packages
npm run build:offline # Rebuild using existing model data without network access
npm run check         # Lint, format, and type check
./test.sh            # Run tests (skips LLM-dependent tests without API keys)
./pi-test.sh         # Run pi from sources (can be run from any directory)

Building standalone binaries from release source

GitHub releases include a versioned source archive covered by the release's SHA256SUMS file. Extract it and run the same build script used for the official standalone binaries:

VERSION="<release-version>"
tar -xzf "pi-${VERSION}-source.tar.gz"
cd "pi-${VERSION}"
./scripts/build-binaries.sh --offline-model-data --platform linux-x64 --out "$PWD/out"

The source archive includes the generated provider model data used for the release. --offline-model-data builds with that snapshot instead of refreshing it from live provider catalogs. The script still installs dependencies, builds the monorepo, compiles the Bun executable, and stages its runtime assets. Package maintainers who provide dependencies separately can pass --skip-install --skip-deps.

Supply-chain hardening

We treat npm dependency changes as reviewed code changes.

• Direct external dependencies are pinned to exact versions. Internal workspace packages remain version-ranged.
• .npmrc sets save-exact=true and min-release-age=2 to avoid same-day dependency releases during npm resolution.
• package-lock.json is the dependency ground truth. Pre-commit blocks accidental lockfile commits unless PI_ALLOW_LOCKFILE_CHANGE=1 is set.
• npm run check verifies pinned direct deps, native TypeScript import compatibility, and the generated coding-agent shrinkwrap.
• The published CLI package includes packages/coding-agent/npm-shrinkwrap.json, generated from the root lockfile, to pin transitive deps for npm users.
• Release smoke tests use npm run release:local to build, pack, and create isolated npm and Bun installs outside the repo before tagging a release.
• Local release installs, documented npm installs, and pi update --self use --ignore-scripts where supported.
• CI installs with npm ci --ignore-scripts, and a scheduled GitHub workflow runs npm audit --omit=dev plus npm audit signatures --omit=dev.
• Shrinkwrap generation has an explicit allowlist for dependency lifecycle scripts; new lifecycle-script deps fail checks until reviewed.

Share your OSS coding agent sessions

If you use Pi or other coding agents for open source work, please share your sessions.

Public OSS session data helps improve coding agents with real-world tasks, tool use, failures, and fixes instead of toy benchmarks.

For the full explanation, see this post on X.

To publish sessions, use badlogic/pi-share-hf. Read its README.md for setup instructions. All you need is a Hugging Face account, the Hugging Face CLI, and pi-share-hf.

You can also watch this video, where I show how I publish my pi-mono sessions.

I regularly publish my own pi-mono work sessions here:

• badlogicgames/pi-mono on Hugging Face

License

MIT

pi.dev domain graciously donated by


exe.dev