Connect over MCP
The alpha tools are for Plus and above. A token carries read, or read and propose; alpha_propose_pod needs propose. Create one while signed in:
curl -X POST https://api.uselyra.dev/api/v2/mcp/tokens -H "Authorization: Bearer <session>" -H "Content-Type: application/json" -d '{"name": "Claude Code", "scopes": ["read", "propose"]}'
The response carries the token once. Keep it out of files you commit.
Claude Code
claude mcp add --transport http lyra https://api.uselyra.dev/mcp --header "Authorization: Bearer lyra_pat_..."
Other clients
Streamable HTTP at https://api.uselyra.dev/mcp with the header Authorization: Bearer <token>.
Tools
| Tool | What it does |
|---|---|
alpha_validate |
Parse an alpha and check it is point-in-time: every input known by the time it trades. Returns the canonical expression, its JSON tree, size and look-back, or the error with its column. Free and unlimited. alpha holds family (a lowercase slug), expression (text or JSON tree), universe {name, instruments, groups}, settings {delay, decay, neutralization, rebalance, truncation} and optional weights. |
alpha_simulate |
Backtest an alpha on Lyra's servers and return the standard output: grade, report, submission checks, breadth and, for books of 5 or more names, a curve of at most 250 points starting at 1. Uses one daily check unless the same spec ran today. Every distinct version counts as a trial of its family. Nothing is saved. alpha holds family (a lowercase slug), expression (text or JSON tree), universe {name, instruments, groups}, settings {delay, decay, neutralization, rebalance, truncation} and optional weights. |
alpha_submit |
Run the submission checks on an alpha and save it to the user's library as the next version of its family when every check passes. Returns the output with saved. A spec simulated today costs no extra check. alpha holds family (a lowercase slug), expression (text or JSON tree), universe {name, instruments, groups}, settings {delay, decay, neutralization, rebalance, truncation} and optional weights. |
alpha_list |
The user's saved alphas, oldest first: id, family, version, expression, grade, passed, sharpe, turnover. family filters to one family; empty lists all. |
alpha_get |
One saved alpha by id: the standard output and the rule as saved. |
alpha_propose_pod |
Propose a pod that trades saved alphas. One id proposes that alpha; 2 to 8 ids blend them at equal risk into a new library entry under family, which uses one daily check. The proposal card lands in the user's Lyra chat. It opens nothing; only the user's Start in the app opens the pod. |
docs_search |
Search the Lyra Labs docs, platform and alpha pages, with a plain question. Returns up to 10 sections, best first: page, title, heading and a line of context. |
docs_get |
One page of the Lyra Labs docs as Markdown with its front matter. Platform pages: overview, getting-started, paper-and-live, chat, research, ideas, your-research, grading, telegram, pods, rules, loss-levels, risk-agent, watch, book, paper-fund, hyperliquid, alpaca, data-engine, fields, custom-data-engine, alpha-playground, ultra, mcp, mcp-tools, security, for-agents, plans, faq, glossary. Alpha pages: alpha/connect, alpha/grammar, alpha/operators, alpha/data, alpha/settings, alpha/output, alpha/checks, alpha/examples. |
data_fields |
The field catalog: each field's syntax, meaning, unit, scope, when it is known, its lag, whether delay 0 can use it, its coverage and the first stored date. No values. |
Checks a day
| Plan | Checks |
|---|---|
| plus | 30 |
| pro | 200 |
| ultra | 200 |
Docs for agents
https://docs.uselyra.dev/llms.txt lists every page. https://docs.uselyra.dev/llms-full.txt is all pages in one file. The Claude Code skill is at https://docs.uselyra.dev/alpha/skill/SKILL.md.