GCFThe AI-native wire format for structured data.

50-92% fewer tokens than JSON. 100% comprehension on every frontier model. 43 billion+ lossless round-trips across 5 formats (JSON, YAML, TOML, CSV, MessagePack). Spec v3.2 Stable.

Five data formats flowing through GCF into the LLM context window

Get Started Try the Playground Cost Calculator GCF vs TOON

Adopted by

·♦️ OmniRoute 6.1K ★·🦞 NetClaw 556 ★·ctx 510 ★·

SDK (A Linux Foundation Project)·🧠 NeuroNest·and more

Zero-code option for MCP servers

One command. 50-92% fewer tokens. Zero code changes.

Zero-code adoption

Drop-in proxy for existing MCP servers. Or integrate natively with any of 6 languages below.

View on GitHub →

pip install gcf-proxyCopy

npm i -g @blackwell-systems/gcf-proxyCopy

go install github.com/blackwell-systems/gcf-proxy@latestCopy

VS Code

Syntax highlighting for .gcf files

Claude Code

One-command proxy setup, session stats

Codex CLI

One-command proxy setup, session stats

n8n

Encode/decode node for workflows

JetBrains

IntelliJ, PyCharm, WebStorm, GoLand

Zed

Tree-sitter syntax highlighting

Zero runtime dependencies · No transitive deps · No supply chain risk · Permanent commitment · MIT licensed

🔄

The universal pivot for structured data

JSON, YAML, TOML, CSV, MessagePack: any format in, GCF in the context window, any format out. One format that speaks every format. No other wire format operates across format boundaries.

decode(encode(value)) == value for every structured value. Verified across 43,000,000,000+ lossless round-trips in 5 formats and 6 languages.

📉

50-92% fewer input tokens

50-69% on a single call (generic to graph profile). Up to 92% with session deduplication across repeated tool calls.

At 1000 orders, JSON doesn't even fit in a 200K context window. GCF fits in 47K. Positional fields, inline schemas, and hierarchical grouping eliminate per-record overhead.

✍️

63% fewer output tokens

LLMs produce valid GCF with a 3-line primer.

33% smaller output than TOON. 5/5 generation validity on every frontier model across 3 providers. Zero training.

🧠

100% comprehension on every frontier model

The only format that never fails. Tested across Claude, GPT-5.5, and Gemini with zero format instructions. On structurally complex code graphs, GCF scores 91.2% where JSON drops to 53.4% and TOON to 68.2%. 2,500+ evaluations, 4 providers.

🔒

Proven lossless. 6 languages. Spec v3.2 Stable.

43B+ (yes, really) round-trips across JSON, YAML, TOML, CSV, and MessagePack. Zero failures. 174 conformance fixtures. Cross-language 6x6 encode/decode matrix verified. Read the spec.

🏆

29% fewer tokens than TOON (16 datasets)

Wins 15/16 real-world datasets. 38% fewer on semi-uniform data, 33% on nested, 32% on K8s pod data. TOON's one win is 77 tokens on a single dataset.

See the Difference

Same data. Fewer tokens. Zero information loss.

JSON458 tokens

{
  "orders": [
    {"id": 1001, "customer": "Acme Corp",
     "total": 49.99, "status": "shipped",
     "items": 1},
    {"id": 1002, "customer": "Globex Inc",
     "total": 150.49, "status": "pending",
     "items": 2},
    {"id": 1003, "customer": "Initech LLC",
     "total": 250.99, "status": "processing",
     "items": 3},

    {"id": 1004, ...},
    {"id": 1005, ...},
    {"id": 1006, ...},
    {"id": 1007, ...},
    {"id": 1008, ...},
    {"id": 1009, ...},
    {"id": 1010, ...}
  ]
}

GCF177 tokens

GCF profile=generic
## orders [10]{id,customer,total,status,items}
1001|Acme Corp|49.99|shipped|1
1002|Globex Inc|150.49|pending|2
1003|Initech LLC|250.99|processing|3
1004|Umbrella Co|351.49|delivered|4
1005|Stark Ind|451.99|shipped|5
1006|Wayne Ent|552.49|pending|6
1007|Oscorp|652.99|shipped|7
1008|LexCorp|753.49|processing|8
1009|Cyberdyne|853.99|delivered|9
1010|Soylent|954.49|shipped|10

61% fewer tokens. Scales to 71%+ at production sizes.

10 rows, 5 fields. Token counts verified with tiktoken (cl100k).

Two lines. Any language.

from gcf import encode_generic, decode_generic

gcf_string = encode_generic(data)
original   = decode_generic(gcf_string)

Two Profiles. One Format.

The generic profile is a strict subset of the graph profile. Call encode(), the LLM reads it natively, call decode() at the end.

Graph-shaped data is the fastest-growing data shape in AI: knowledge systems, ontologies, GraphRAG, code intelligence, agent memory. No other token-efficient format treats graphs as first-class. GCF is the only format with native graph syntax: local IDs, typed edges, distance grouping, and session deduplication that compounds to 92% savings across multi-turn sessions.

Generic Profile

Any structured data (subset)

Any structured value in, same value out. Verified lossless across 43 billion+ round-trips with JSON, YAML, TOML, CSV, and MessagePack. 71% fewer tokens than JSON.

JSON

[
  {"name":"validateToken",
   "kind":"func","refs":18},
  {"name":"refreshSession",
   "kind":"func","refs":6},
  {"name":"getConnection",
   "kind":"func","refs":34},
  {"name":"runMigration",
   "kind":"func","refs":3}
]

GCF

## results [4]{name,kind,refs}
validateToken|func|18
refreshSession|func|6
getConnection|func|34
runMigration|func|3

Graph Profile

Superset: adds IDs, edges, scores

Knowledge graphs, code intelligence, ontologies, relationship networks. Graph-shaped data is the fastest-growing data shape in AI. No other format treats it as a first-class citizen.

JSON

{
  "symbols": [
    {"id":1,"kind":"func",
     "name":"handleReq"},
    {"id":2,"kind":"func",
     "name":"validate"},
    {"id":3,"kind":"iface",
     "name":"AuthCfg"}
  ],
  "edges": [
    {"src":1,"tgt":2,
     "type":"calls"},
    {"src":2,"tgt":3,
     "type":"implements"}
  ]
}

GCF

## symbols [3]
@1 func handleReq 0.95
@2 func validate 0.87
@3 iface AuthCfg 0.60

## edges [2]
@2<@1 calls
@3<@2 implements

Session Dedup

Graph profile

JSON retransmits everything on every call. GCF tracks which symbols have been sent and only transmits bare references for known ones. 92% savings by the 5th call.

JSON (call 2: full retransmit)

[
  {"id":1,"kind":"func",
   "name":"handleReq"},
  {"id":2,"kind":"func",
   "name":"validate"},
  {"id":3,"kind":"iface",
   "name":"AuthCfg"},
  {"id":4,"kind":"func",
   "name":"revoke"}
]

GCF (call 2: bare refs + new)

## symbols [4]
@1
@2
@3
@4 func revoke 0.91

GCF Grammar

Two profiles. Session-aware. No ambiguity.

Generic Profile

Section Headers

## symbols [4]{kind,qname}
## edges [3]
## targets
##! summary

## section start
[N] element count
{fields} inline schema
[?] deferred count (streaming)
##! summary trailer

Tabular Data

## users [3]{name,role,active}
Alice|admin|true
Bob|dev|true
Carol|dev|false

| pipe-separated values
Fields declared once in header
No quotes unless needed
No braces, no colons per row

Scalars & Key-Value

name=Alice
age=30
active=true
missing=-
empty=""

key=value for primitives
- null, ~ absent, "" empty string
Quote if value contains | or newline

Nested Objects

## orders [2]{id,total,"customer>name","customer>tier"}
1001|249.99|Alice|premium
1002|89.50|Bob|standard

"customer>name" flattens nested field into column
> separates path levels
Values go directly in the row (no attachments)

Graph Profile

Symbols

@1 func handleReq 0.95
@2 iface AuthCfg 0.60 ast
@3
@4 func revoke 0.91

@N local ID, kind, qname, score, provenance
Bare @N = session ref (already transmitted)

Edges

@2<@1 calls
@3<@1 calls
@4<@2 implements

@target<@source type
One edge per line, no nesting overhead

Distance Groups

## targets [3]
@0 fn handleReq 0.95 lsp
@1 fn validate 0.87 lsp
@2 fn connect 0.91 lsp

## related [2]
@3 fn helper 0.60 ast
@4 iface Config 0.55 ast

targets, related, extended by relevance
LLM reads count from header, no scanning

Session Dedup graph profile

GCF profile=graph tool=blast_radius symbols=6 session=true
## targets
@0  # previously transmitted
@1  # previously transmitted
@5 fn pkg.NewFunc 0.85 lsp

session=true signals bare refs present
@N # previously transmitted = sent in prior call
LLM already has the full declaration in context
88% savings vs JSON across multi-turn sessions

Delta Encoding graph profile

GCF tool=topology delta=true tokens=30 savings=85%
## removed
fn pkg.OldHandler
## added
@0 fn pkg.NewHandler 0.85 lsp
## edges_added
pkg.Router -> pkg.NewHandler calls

delta=true signals diff from prior payload
Only added/removed symbols and edges
95% savings for small topology changes

Streaming both profiles

## results [?]{name,kind}
validate|func
connect|func
handle|func
##! summary counts=3

[?] count unknown upfront
Rows emit instantly, O(1) memory
##! trailer finalizes count
Zero buffering, zero latency