Platform builders composing agentic products need focused, interoperable libraries and services — not monolithic frameworks. Gear provides individual capabilities that work standalone or together, with humans and agents sharing the same interface and documentation.
What becomes possible
For Platform Builders
Give humans and agents shared capabilities through the same
interface, with tooling to prove changes improved outcomes. Every
CLI prints grep-friendly help and JSON output; every library ships a
matching skill in the forwardimpact/fit-skills pack so
agents land on the same docs as humans.
The pack ships these packages, all published to npm under
@forwardimpact/lib* and
@forwardimpact/svc*:
- 40 libraries
- 15 services
Browse the full tables in libraries/README.md and services/README.md.
Bridge Threaded Channels to the Agent Team
Ship a chat or discussion adapter without rebuilding the intake skeleton, callback registry, and durable thread state from scratch each time.
Enable Agents on Every Surface
Give agents and humans the same interface so capabilities ship once. Render structured output across web and terminal from shared handler logic.
Ground Agents in Context
Answer relationship questions, look up context fast, and give agents typed, retrievable knowledge with semantic search — no external database required.
Ground Service Contracts in One Source
Publish a service proto that resolves on any external install, so consumers never chase a shared schema living in a package they never pulled in.
Integrate with the Engineering Standard
Distribute skill packs through the tools agents and engineers already use, and turn engineering standard definitions into queryable, derivable data.
Keep Instruction Layers Honest
Enforce instruction-layer length caps and JTBD invariants automatically, so a docs change cannot quietly drift the layered architecture.
Keep Service Contracts Typed
Keep JavaScript types in sync with proto definitions and register gRPC services as MCP tools from config, with no hand-written glue.
Run a Predictable Platform
Preflight checks before anything heavy runs, supervised processes, structured telemetry, and instruction layers that stay honest.
Prove Agent Changes
Get reproducible evidence for whether an agent change actually helped, and generate complete eval datasets from a single DSL file.
Ship Predictable CLIs
Enforce a runtime floor before any heavy import evaluates, so a CLI surfaces your version error instead of a confusing one from deep in a dependency.
Ship Predictable Services
Fail fast at startup before a service constructs partially against an unsafe default, turning a silent misconfiguration into a clear early error.
Ship Service Endpoints Without Boilerplate
Stand up an HTTP or gRPC endpoint without reimplementing server lifecycle, security headers, health checks, or transport every time.
For Empowered Engineers
Run an agent team that remembers what it learned and acts on real signal. Give the team memory that survives across sessions, and separate a genuine shift in a metric from ordinary variation before anyone reacts to it.
Getting Started
npm install @forwardimpact/libcli @forwardimpact/libstorage # any subset
apm install forwardimpact/fit-skills