langchain-router
Your agent doesn't need the expensive model for every call.
Most calls are just the model picking which file to read next or which pattern to search for. A smaller model does that fine. This middleware detects when the agent is doing that kind of work and routes to a fast model automatically.
Quick Install
pip install langchain-router
🤔 What is this?
Agent sessions have a pattern. The user says something, the agent thinks about it (planning). Then it reads files, searches code, runs commands (execution). Sometimes something breaks (recovery). Then the user says something again.
Planning and recovery need the primary model. Execution doesn't. RouterMiddleware detects which phase the agent is in and routes accordingly.
| What just happened | Phase | Model |
|---|---|---|
| User spoke | planning | primary |
| Tool call succeeded | execution | fast |
| Tool call failed | recovery | primary |
No ML. No embeddings. No classifier. Just message types.
from langchain.agents import create_agent
from langchain_router import RouterMiddleware
agent = create_agent(
model="anthropic:claude-sonnet-4-6",
tools=[...],
middleware=[RouterMiddleware(fast="anthropic:claude-haiku-4-5-20251001")],
)
With CollapseMiddleware
from langchain_collapse import CollapseMiddleware
middleware = [
CollapseMiddleware(),
RouterMiddleware(fast="anthropic:claude-haiku-4-5-20251001"),
]
flowchart TB
A["📥 37 messages"] --> B["CollapseMiddleware"]
B --> C["📥 9 messages"]
C --> D["RouterMiddleware"]
D --> E{"phase?"}
E --> |"execution · 83%"| F["⚡ GLM-5"]
E --> |"planning"| G["🧠 Opus 4.6"]
E --> |"recovery"| G
style A fill:#ff6b6b,stroke:#e03131,color:#fff
style B fill:#339af0,stroke:#1c7ed6,color:#fff
style C fill:#339af0,stroke:#1c7ed6,color:#fff
style D fill:#51cf66,stroke:#2f9e44,color:#fff
style E fill:#fff3bf,stroke:#f59f00,color:#333
style F fill:#20c997,stroke:#099268,color:#fff
style G fill:#845ef7,stroke:#7048e8,color:#fff
Verified with real API calls
81 real API calls across 9 model pairs, 4 providers. Each pair ran a 9-call agent session: planning, 6 execution turns, recovery from a failed tool call, and re-planning. Every call routed correctly. Full per-call details.
| Primary | Fast | Provider | Calls | Routing |
|---|---|---|---|---|
| Claude Opus 4.6 | Claude Haiku 4.5 | Anthropic → Anthropic | 9 | 9/9 ✓ |
| Claude Opus 4.6 | GLM-5 | Anthropic → Z.ai | 9 | 9/9 ✓ |
| Claude Opus 4.6 | MiniMax M2.7 | Anthropic → MiniMax | 9 | 9/9 ✓ |
| Claude Sonnet 4.6 | GLM-5 | Anthropic → Z.ai | 9 | 9/9 ✓ |
| Claude Sonnet 4.6 | MiniMax M2.7 | Anthropic → MiniMax | 9 | 9/9 ✓ |
| GPT-5.4 | GPT-4.1-mini | OpenAI → OpenAI | 9 | 9/9 ✓ |
| GPT-5.4 | GLM-5 | OpenAI → Z.ai | 9 | 9/9 ✓ |
| GPT-5.4 | MiniMax M2.7 | OpenAI → MiniMax | 9 | 9/9 ✓ |
| GLM-5 | MiniMax M2.7 | Z.ai → MiniMax | 9 | 9/9 ✓ |
Patterns tested: same-provider (Anthropic→Anthropic, OpenAI→OpenAI), cross-provider (Anthropic→Z.ai, OpenAI→MiniMax), frontier→open (Opus→GLM-5, GPT-5.4→MiniMax), open→open (GLM-5→MiniMax).
Cost impact
In a typical coding session (8 file reads, 4 greps, 1 edit, 1 test failure, 1 fix, 1 re-test, 2 user messages), 83% of model calls are execution phase and route to the fast model.
| Primary | Fast | Per session | Saved | Annual (10 devs, 20/day) |
|---|---|---|---|---|
| Opus 4.6 | MiniMax M2.7 | $4.19 → $0.75 | 82% | $171,938/yr |
| Opus 4.6 | GLM-5 | $4.19 → $0.80 | 81% | $169,125/yr |
| Opus 4.6 | Haiku 4.5 | $4.19 → $0.88 | 79% | $165,075/yr |
| Sonnet 4.6 | MiniMax M2.7 | $0.84 → $0.19 | 78% | $32,437/yr |
| GPT-5.4 | MiniMax M2.7 | $0.77 → $0.18 | 77% | $29,438/yr |
| GPT-5.4 | GPT-4.1-mini | $0.77 → $0.21 | 72% | $27,675/yr |
| Sonnet 4.6 | GLM-5 | $0.84 → $0.24 | 71% | $29,625/yr |
| GPT-5.4 | GLM-5 | $0.77 → $0.23 | 70% | $26,625/yr |
| GLM-5 | MiniMax M2.7 | $0.13 → $0.07 | 45% | $2,813/yr |
Token counts assumed for cost projection (8k input, 1.5k output per call). Pricing as of April 2026. See benchmark for methodology.
On false positives
Recovery detection checks ToolMessage.status first (set by the agent when a tool call fails), then falls back to content scanning for error, traceback, exception, failed. Code containing those words (like def handle_error) routes to the primary model. That's the safe direction: more capability than needed, never less.
📖 Documentation
- Source (single file, ~170 lines)
- Benchmark (cost projections across providers)
- Verified Results (81 real API calls, 9 model pairs)
- Tests (unit, integration, and property based invariant tests)
💁 Contributing
git clone https://github.com/johanity/langchain-router.git
cd langchain-router
pip install -e ".[test]"
pytest
📕 License
MIT
Metadata
Release files for langchain-router 0.1.4
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | Provenance | Actions |
|---|---|---|---|---|
| langchain_router-0.1.4.tar.gz | 16.2 kB | Without provenance | Details for langchain_router-0.1.4.tar.gz |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Actions Reset |
|---|---|---|---|---|
| langchain_router-0.1.4-py3-none-any.whl | Python 3 | none | any | Details for langchain_router-0.1.4-py3-none-any.whl |
Total release size: 22.8 kB
Release files / langchain_router-0.1.4.tar.gz
| Download URL | langchain_router-0.1.4.tar.gz |
|---|---|
| Size | 16.2 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
fca3912cfd501a77940a470f73678a58cd0223698edd730dd37eab2622c7e7b3
|
|
BLAKE2b-256 checksum How to use checksums |
05f9f9d62c48238593bd440b14981ba3b54a448e1eef475eab623e8eb3b2c9dc
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/6.2.0 CPython/3.12.12
|
Release files / langchain_router-0.1.4-py3-none-any.whl
| Download URL | langchain_router-0.1.4-py3-none-any.whl |
|---|---|
| Size | 6.6 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
2c29768bcff019170fb4d58c76f2f1ddafbf581323ec07a72cebf95e6efd6dd9
|
|
BLAKE2b-256 checksum How to use checksums |
89867690cda113e41f2abe6ebb19814410b9255b125e1fb9289655821a2cbbc0
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/6.2.0 CPython/3.12.12
|
Security for langchain-router 0.1.4
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
No provenance found
None of the files in this release were uploaded with provenance metadata.
A note on package security
PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.
To learn more about protecting yourself from malicious actors, visit our package security documentation.