@blackwell-systems/gcf weekly downloads statistics

gcf-typescript

TypeScript implementation of GCF, the most token-efficient wire format for LLMs. A drop-in alternative to JSON and TOON for any structured data.

Built for the agentic loop, where the same structured context crosses the model boundary turn after turn. A single payload is 50-92% smaller than JSON, but GCF also deduplicates repeated structure across turns and sends only deltas when context changes, so by the 5th overlapping call each response costs 99% fewer tokens than JSON, and a 10-call session runs 94.4% cheaper than re-sending JSON every turn. Session dedup and delta both need local IDs and a multi-turn design that neither JSON nor TOON has.

• 100% comprehension on every frontier model, zero training. 29% fewer tokens than TOON and 56% fewer than JSON across 16 datasets; 91.2% on structurally complex code graphs (vs TOON 68.8%, JSON 54.1%). • Proven lossless across 43,000,000,000+ round-trips in 5 formats and 6 languages. Zero runtime dependencies. • One format, four properties no other single format holds at once: schema-free, lossless, token-compact (50-92% vs JSON), and model-readable with zero training. JSON is verbose, Protobuf needs a schema, MessagePack is binary, and TOON isn't reliably lossless.

2,500+ LLM evaluations. Full benchmarks.

Docs: gcformat.com · Playground · GCF vs TOON

Install

Zero dependencies. TypeScript-first. Includes CLI. Don't want to change code? Use the MCP proxy for zero-code adoption.

CLI

Or install globally: npm install -g @blackwell-systems/gcf then use gcf directly.

Library

Quick Start

Output:

Decode

Session Deduplication

Track transmitted symbols across multiple tool responses. Previously-sent symbols become bare references instead of full declarations:

By the 5th call in a session: 86% fewer tokens than JSON from dedup alone, 99% stacked with delta encoding.

Streaming Encode

Write GCF output incrementally as symbols and edges arrive. Zero buffering, O(1) memory per row. Ideal for MCP servers that walk large graphs or paginate results:

Output:

The writer is any object with a write(s: string) method (Node.js streams, web WritableStreams, or a simple callback). Standard decode() handles streaming output with no changes.

Delta Encoding

When the consumer already has a prior context pack, send only what changed:

81.2% savings on re-queries where the pack changed slightly.

Generic Encoding

Encode any JS value (not just graph payloads) into GCF tabular format:

Output:

Works on objects, arrays, and primitives. Arrays of uniform objects get tabular rows. Nested objects use ## key section headers.

Generic-Profile Delta (multi-turn)

In an agent loop the same keyed table gets re-queried turn after turn. Instead of re-sending the whole table each time, send only the changed rows (SPEC §10a):

Opt-in and bilateral, keyed on content-addressed pack roots. By the 5th overlapping call, ~97% fewer tokens than re-sending JSON.

Re-anchor session helper

GenericDeltaSession manages the delta/re-anchor cadence for you: each next() returns either a compact delta or, on its cadence, a full re-anchor (which re-grounds the consumer), updating its held base.

fixedN(15) re-anchors every N turns; sizeGuard() (recommended) re-anchors once the cumulative delta reaches a full payload's size. It introduces no new wire syntax and the decoder stays cadence-agnostic, so a re-anchor is just the protocol's "full" outcome on a schedule.

API

| Function | Description |

| encode(p: Payload): string | Encode a graph payload to GCF text | | encodeGeneric(data: unknown): string | Encode any value to GCF tabular format | | decode(input: string): Payload | Parse GCF text back to a Payload | | encodeWithSession(p: Payload, s: Session): string | Encode with session deduplication | | new StreamEncoder(w, tool, opts) | Create a streaming encoder (zero-buffering) | | encodeDelta(d: DeltaPayload): string | Encode a delta (added/removed only) | | diffGenericSets(base, next): GenericDeltaPayload | Diff two keyed record sets (generic profile) | | encodeGenericDelta(d): string / decodeGenericDelta(s) | Generic-profile delta wire (§10a) | | verifyGenericDelta(base, d, root): GenericSet | Atomic apply + new_root verification | | new GenericDeltaSession(base, tool, policy) | Producer-side re-anchor cadence helper (§10a.8) | | new Session() | Create a new session tracker |

Types

| Type | Purpose |

| Payload | Full GCF payload: tool, budget, symbols, edges, pack root | | Symbol | Graph node: qualified name, kind, score, provenance, distance | | Edge | Directed relationship: source, target, edge type | | DeltaPayload | Diff between two packs: added/removed symbols and edges | | GenericSet / GenericDeltaPayload | Keyed record set and its generic-profile diff (§10a) | | GenericDeltaSession | Stateful producer that schedules delta vs full re-anchor (§10a.8) | | Session | Tracker for multi-call deduplication | | KIND_ABBREV / KIND_EXPAND | Bidirectional kind abbreviation maps |

Benchmarks

2,500+ LLM evaluations across 11 models, 4 providers, and 50+ independent test runs.

| | GCF | TOON | JSON |

| Comprehension (23 runs, 10 models) | 91.2% | 68.8% | 54.1% | | Generation (28 runs, 9 models) | 5/5 | 1.0/5 | 5.0/5 | | Input tokens (500 symbols) | 11,090 | 16,378 | 53,341 | | Output tokens (100 symbols) | 5,976 | 8,937 | 16,121 |

GCF wins 15/16 datasets on the expanded token efficiency benchmark. Full results: gcformat.com/guide/benchmarks

Implementations

| Language | Package | Repository |…