Harvest losses without the wash
Find the lots, pick replacements that keep exposure, respect the 61-day window across every account.
Get startedAbacus Labs makes portfolio decisions as reliable and programmable as APIs
Built for the agents people already use
From a sentence to a checked trade list, Abacus Labs powers agents that manage money reliably.
Portfolio options, built for agents. Send an account and a goal in plain English, get back 2 to 5 options with their tradeoffs.
Try for freeGive your agent a rule checker. Send the trades it drafted to propose instead of a goal, get back the violations and a repaired list that a broker will accept.
Try for freeWhat is this person exposed to, and how much risk are they taking? Ask in plain English, or call the components directly: holdings, exposures, factor exposures, risk, tracking error, tax and drift, each with structured numbers and a plain-English note.
Try for freeThe portfolio is the person's native interface to the market.
Now your agents can build one, from a sentence, with every rule checked.
Implement the goal. Explain the tradeoff. Never place the order.
Find the lots, pick replacements that keep exposure, respect the 61-day window across every account.
Get startedTrack a target, trade only what drifted, use losses to pay for it.
Get startedA theme, a tilt, a conviction list. Abacus Labs sizes it against what the person already holds.
Get startedSet a risk level in plain words. Abacus Labs finds the cheapest way there in tax and turnover.
Get startedTrim now, trim with losses, or hedge around it. The person sees all three.
Get startedSend every LLM-drafted trade list to propose. Get the violations and a repaired list.
Get startedSectors, styles, overlap, concentration. Plain English the agent can repeat.
Get startedPut a watch on an account. Abacus Labs runs the analysis components on a schedule and posts to your webhook when something is worth telling the person.
Get startedConnect an account, propose, and commit your first checked trade list
Pass the account inline or by id after linking an aggregator. Sandbox keys come with example accounts.
Your agent shows the person the options and their tradeoffs, and runs what-ifs inside the conversation.
An exact solver re-solves the chosen option and returns the final trade list with a certificate of what was checked.
curl https://api.abacuslabs.ai/v1/propose \
-H "Authorization: Bearer $ABACUS_API_KEY" \
-d '{ "account": { "id": "acct_demo_1" },
"goal": "reduce NVDA concentration, keep taxes low",
"options": 3 }'
curl https://api.abacuslabs.ai/v1/commit \
-H "Authorization: Bearer $ABACUS_API_KEY" \
-d '{ "option_id": "opt_7f3a" }'
import os, requests
API = "https://api.abacuslabs.ai/v1"
H = {"Authorization": f"Bearer {os.environ['ABACUS_API_KEY']}"}
opts = requests.post(f"{API}/propose", headers=H, json={
"account": {"id": "acct_demo_1"},
"goal": "reduce NVDA concentration, keep taxes low",
"options": 3}).json()
plan = requests.post(f"{API}/commit", headers=H, # exact solve + certificate
json={"option_id": opts["options"][0]["id"]}).json()
{ "mcpServers": { "abacus": {
"url": "https://mcp.abacuslabs.ai",
"headers": { "Authorization": "Bearer ABACUS_API_KEY" } } } }
No fractional shares in a whole-share account.
No wash sales. No cash that does not add up.
Pay for decisions, not tokens. Every plan includes portfolios Abacus Labs watches for you and a monthly pool of credits.
Sign in with Google, GitHub, Microsoft, an email link or a passkey. The Personal plan is free: 3 portfolios and 300 credits a month, no card.
Building with an agent? Hand it abacuslabs.ai/agent.md and it can self-serve.
Your key shows once, with copy-paste blocks for curl, Python and MCP. The dashboard waits for your first call and shows the response.