providers.example.txt

# =============================================================================
# UNIFIED LLM PROVIDER CONFIGURATION REFERENCE
# =============================================================================
# Format:    AGENTS STANDARD PROVIDER REFERENCE v2.0.0
# Location:  ~/.agents/providers.txt
# Purpose:   Master configuration reference for ALL agentic tools.
#            kilo, crush, pi, goose, dcode, hermes, cline, qwen-code, Zed,
#            Agy/Gemini CLI, Codex, Windsurf, Cursor, Copilot, Junie, and more.
#
# This file defines LLM providers and models in a single, agent-agnostic format
# that any coding agent can read. Tool makers retain their own per-tool configs
# for provider/model overrides — this is the base standard, not a replacement.
#
# SCHEMA: providers.txt is a tab-separated-value (TSV) reference document
# organized into sections delimited by `# ===...===` headers. The file is both
# human-readable documentation AND machine-parseable configuration.
#
# SECURITY: Provider API keys are NEVER stored in this file. Load keys through
#           secure runtime environments, PQC-encrypted bundles, or agent-native
#           credential stores. Use ${ENV_VAR} references only.
# =============================================================================

# =============================================================================
# PROVIDER SUMMARY TABLE
# =============================================================================
# Format: │ Provider-Namespace │ API Endpoint                          │ Key Env Var    │ Default Agent │
#
# The Provider-Namespace is the canonical key used throughout this file
# and by agents referencing it. Use lowercase, hyphen-separated names.
#
# Example entries (replace with your own providers):
#
# │ your-provider      │ https://api.your-provider.com/v1            │ YOUR_API_KEY   │ —            │
# │ openai             │ https://api.openai.com/v1                   │ OPENAI_API_KEY │ —            │
# │ anthropic          │ https://api.anthropic.com/v1                │ ANTHROPIC_API_KEY│ —          │
# │ local              │ http://localhost:11434/v1                   │ —              │ —            │
#
# For each provider, document: namespace, endpoint URL, required env var,
# and which agent uses it as default (if any).

# =============================================================================
# MASTER MODEL SPECIFICATION
# =============================================================================
# Format (TSV, tab-separated):
# │ # │ Provider           │ Model ID (API)                │ Display Name          │ Context  │ Output   │ Tools │ Img  │ Cache │ Reasoning │
#
# COLUMN DEFINITIONS:
#   #           — Row number (sequential, for human reference)
#   Provider    — Provider namespace from the summary table above
#   Model ID    — Exact API model identifier (as sent in the `model` field)
#   Display     — Human-readable display name (used in agent UIs/pickers)
#   Context     — Maximum context window in tokens
#   Output      — Maximum output tokens
#   Tools       — YES/NO: supports tool calling / function calling
#   Img         — YES/NO: supports image/multimodal input
#   Cache       — YES/NO: supports prompt caching
#   Reasoning   — YES/NO/NO*: supports native reasoning (NO* = supported but
#                 deliberately disabled for compatibility)
#
# Example entries:
#
# │ 1 │ openai             │ gpt-5.4                       │ openai:gpt-5.4           │ 200,000  │ 128,000  │ YES   │ YES  │ YES   │ NO        │
# │ 2 │ openai             │ o4-mini                       │ openai:o4-mini           │ 200,000  │ 100,000  │ YES   │ YES  │ NO    │ YES       │
# │ 3 │ anthropic          │ claude-opus-4-7               │ anth:claude-opus-4-7     │ 200,000  │ 32,768   │ YES   │ YES  │ YES   │ NO        │
# │ 4 │ anthropic          │ claude-sonnet-4-6             │ anth:claude-sonnet-4-6   │ 200,000  │ 32,768   │ YES   │ YES  │ YES   │ NO        │
# │ 5 │ local              │ llama-4-maverick              │ local:llama-4            │ 131,072  │ 8,192    │ YES   │ YES  │ NO    │ NO        │
#
# DISPLAY NAME CONVENTION: <provider-short-prefix>:<model-short-name>
#   Use a 2-5 character prefix that uniquely identifies the provider.
#   The model short name should be readable and distinguishable.

# =============================================================================
# PROVIDER DETAILS
# =============================================================================

## PROVIDER SECTION FORMAT __________________________________________________
#
# For each provider defined in the summary table, create a detail section:
#
# ## <provider-name> _________________________________________________________
# # Tier:        <service tier / classification>
# # Endpoint:    <API base URL>
# # Key env:     <ENVIRONMENT_VARIABLE_NAME>
# # Prefix:      <2-5 char display prefix>
# # Cross-tool:  <which agents support this provider>
# # Models:      <count> — <comma-separated list of models>
# # Notes:       <freeform notes, caveats, special behavior>
#
# Include model-specific notes like native reasoning behavior,
# caching specifics, multimodal capabilities, and compatibility
# considerations.

## EXAMPLE PROVIDER _________________________________________________________
# Tier:        Example
# Endpoint:    https://api.example.com/v1
# Key env:     EXAMPLE_API_KEY
# Prefix:      ex
# Cross-tool:  All agents (openai_compatible format)
# Models:      2  —  example-model-v1, example-model-v2
# Notes:       This is a provider section template. In a real providers.txt,
#              replace with actual provider documentation.

# =============================================================================
# CROSS-TOOL CONFIGURATION MATRIX
# =============================================================================
# Format:
# │ Provider    │ Agent1 │ Agent2 │ Agent3 │ ... │
#
# Documents which agents support each provider. All providers using the
# standard OpenAI-compatible API format (POST /v1/chat/completions) work
# with any tool that supports custom OpenAI-compatible endpoints.
#
# Example:
#
# │ Provider    │ Hermes │ DeepAgents │ Zed    │ kilo │ crush │ pi  │ goose │ dcode │ cline │ qwen  │
# │-------------+--------+------------+--------+------+-------+-----+-------+-------+-------+-------│
# │ your-prov   │ YES    │ YES        │ custom │ YES  │ YES   │ YES │ YES   │ YES   │ YES   │ YES   │
#
# Notes:
# - "YES" = works out of the box with the agent's native provider config
# - "custom" = requires custom provider definition (e.g., Zed custom provider)
# - "—" = not tested or not supported

# =============================================================================
# FORMAT SPECIFICATION
# =============================================================================
#
# VERSION:        2.0.0
# SCHEMA:         https://agentsstandard.com/schemas/providers.txt
# MIME:           text/plain; charset=utf-8
# ENCODING:       UTF-8
# LINE ENDINGS:   LF (\n)
#
# PARSING RULES:
#   1. Lines starting with `# ` are descriptive comments (human documentation)
#   2. Sections are delimited by `# ===...===` headers
#   3. Provider summary: pipe-delimited table under `# PROVIDER SUMMARY TABLE`
#   4. Model specification: pipe-delimited TSV table under `# MASTER MODEL SPECIFICATION`
#   5. Provider details: `## <name> ___...___` sections with `# Key: Value` fields
#   6. Cross-tool matrix: pipe-delimited table under `# CROSS-TOOL CONFIGURATION MATRIX`
#   7. Numeric values may include commas for readability (e.g., "1,000,000")
#   8. Boolean fields use YES/NO (case-sensitive)
#   9. The `NO*` reasoning value means "supported but deliberately disabled"
#
# AGENT INTEGRATION:
#   Agents should:
#   1. Check for ~/.agents/providers.txt at startup
#   2. Parse provider sections into internal config
#   3. Merge with agent-specific provider overrides
#   4. Use the cross-tool matrix to identify compatible providers
#
#   Tool makers should:
#   1. Read ~/.agents/providers.txt as a base configuration
#   2. Maintain their own per-tool config files for overrides
#   3. Fall back to providers.txt for unknown providers
#   4. Contribute provider updates back to the standard
#
# RELATIONSHIP TO OTHER STANDARDS:
#   - ~/.agents/AGENTS.md     → behavioral rules (how agents should behave)
#   - ~/.agents/mcp-settings.json → MCP server catalog (what tools are available)
#   - ~/.agents/providers.txt     → LLM provider config (which models to use)
#   - ~/.agents/skills/           → Agent skills (what agents can do)
#
# CONTRIBUTING:
#   Submit provider updates, new models, and corrections via:
#   https://github.com/nbiish/agents-standard/issues/new?template=provider-update.md
#
#   See CONTRIBUTING.md for the full contribution protocol.
#   See providers.json for the machine-readable structured provider data.
#
# =============================================================================
# END OF TEMPLATE
# =============================================================================