Marketplace

Frontier model offers and MCP access on Portal compute. Activate in one click and pay in USD through a single Portal wallet - one key, one invoice, every call audited.

Connect over MCP
Portal MCP Live

Connect Portal over the Model Context Protocol, call Portal-branded customer routes from your agent, and browse every marketplace offer and skill as tools. Point Cursor, the Portal SDK, or any MCP client at the endpoint below, add your sk-portal-… key, and your agent runs prompts on Portal compute.

Docs →
◎ Streamable HTTP Call any Portal model All offers as tools All skills as tools Works in Cursor · Portal · any MCP client
Add to your MCP client (e.g. Cursor ~/.cursor/mcp.json)
{ "mcpServers": {
    "portal": {
      "url": "https://api.portal.ai/mcp",
      "headers": { "Authorization": "Bearer sk-portal-…" }
    }
} }

Tools: portal_call (plain text in, the model's reply out, per-token billing; defaults to portal/m2-25, with portal/m3-55-xh-1m for the hardest work) · portal_models_list · portal_offers_list · portal_offer_get · portal_skills_search · portal_skill_get · portal_skill_categories · portal_activate. Discovery is open without a key; portal_call needs your sk-portal-… key in the header above.

Sold by Portal
Portal Anchor seller
Portal-branded customer routes with one stable model id per tier, streaming, tool calls, and prompt caching. Metered per token in USD and billed to your Portal wallet.
Sold by Portal · settled by Portal
Portal M2-25

A frontier agentic coding model built for long-context reasoning, precise edits and tool use - exposed through one stable model id. Non-fast and priority Fast tiers.

Activate →
200K context Market frontier models Streaming Tool / function calling Prompt caching
Model id
portal/m2-25
Context window
200,000 tokens
Tiers
Non-fast · Fast
Billing
Per token, USD
Non-fast · per 1M tokens · USD
Input
$0.12
/ 1M
Cached input
$0.05
/ 1M
Cache write
$0.12
/ 1M
Output
$0.59
/ 1M

Priority portal/m2-25-fast tier available at $0.70 / $0.30 / $0.70 / $3.50 per 1M. See full specs on Models.

Portal M3-55 XH 1M

Frontier general-purpose intelligence at Extra-High reasoning effort with a 1M-token context window. The deepest-thinking tier in the Portal lineup - for complex agents, analysis and tool-heavy workflows.

Activate →
1M context Extra-High reasoning effort Market frontier models Streaming Tool / function calling Prompt caching
Model id
portal/m3-55-xh-1m
Context window
1,000,000 tokens
Reasoning effort
Extra-High
Billing
Per token, USD
Per 1M tokens · USD
InputTokens you send in the request
$3.50
/ 1M
Cached inputCache read - repeated context served from cache
$0.35
/ 1M
Cache writeWriting new context into the cache
$3.50
/ 1M
OutputTokens the model generates
$21.00
/ 1M

Reasoning effort and context length don't change the per-token rate - pick Extra-High for the hardest problems at no premium over the base tier. Pricing is metered per token in USD and billed against your wallet.

Portal M3 500K Fast New

Flagship-speed frontier tier: near-top reasoning at priority speed with a 500K-class context window. The default when you want strong and fast at once.

Activate →
500K context Priority speed Market frontier models Streaming Tool / function calling Prompt caching
Model id
portal/m3-500k-fast
Context window
500,000 tokens
Speed tier
Fast (priority)
Billing
Per token, USD
Per 1M tokens · USD
Input
$2.80
/ 1M
Cached input
$0.70
/ 1M
Cache write
$2.80
/ 1M
Output
$8.40
/ 1M

This marketplace contains Portal-branded customer offers. Underlying routing and availability may change.