PlatformMCP serverNew
Your agent, but it reads filings now.
Add one line to your config. The next time your model asks "what just moved $NVDA?", it answers itself, with the same verdicts a trader pays for.
[01]Compatibility
Any MCP host, same endpoint. One server. The protocol does the rest. Wire it into whatever model interface your team already lives in.
C
Claude Desktop
Native
Cursor
Native
Continue
Plugin
Cline
Plugin
OpenAI Agents
SDK
Custom
SDK
Forecite · MCP server
mcp://forecite.com/v1
spec compliant · v2024.11
[02]Tools
Four tools, all the model needs. Tight surface, JSON-schema-typed inputs, predictable outputs. Models stop hallucinating arguments when the spec is sharp.
search_verdicts()
toolQuery historical verdicts by ticker, sector, or score.
Inputs
- ticker:stringrequiredUppercase symbol, no $ prefix
- since:durationoptionalDefault: 24h
- min_score:floatoptionalRange −1.0 to +1.0
ReturnsPage<Verdict>
score_text()
toolRun arbitrary text through the Verdict Engine.
Inputs
- text:stringrequired
- ticker:stringoptionalHint for entity disambiguation
ReturnsVerdict
fetch_filing()
toolPull a SEC filing by accession number; returns parsed body.
Inputs
- accession:stringrequired
- format:'text' | 'html'optionalDefault: text
ReturnsFiling
watch_ticker()
toolSubscribe an agent loop to live verdicts for a ticker.
Inputs
- ticker:stringrequired
- max_events:intoptionalDefaults to unbounded
ReturnsAsyncIter<Verdict>
[03]A turn in the agent
What it looks like, mid-conversation. Tools are called inline. Results stay structured. The model writes back grounded in numbers, not summaries.
Claude · with Forecitesession 04a91f
What just moved $NVDA in the last hour, and how confident is your read?
→ search_verdicts(ticker="NVDA", since="1h", min_score=0.6)
← search_verdicts · 2 verdicts → +0.83 (8-K beat), +0.41 (guidance reiteration)
Two strong-positive verdicts in the last hour: an 8-K earnings beat (verdict +0.83, conf 0.91) at 16:01:03 UTC and a guidance reiteration (+0.41, conf 0.74) at 16:02:14. The first is the mover. The second is post-call confirmation. Want the underlying filing text?
[04]Resources & prompts
More than tools. Resources let the model read the world; prompts give it starter scaffolds. Both are addressable by URI: no chatter, no re-prompting.
mcp://forecite.com/v1
verdicts/
by-ticker/{symbol}/latestVerdict · latest snapshot
by-ticker/{symbol}/window/{n}hPage<Verdict>
by-sector/{slug}/todayPage<Verdict>
filings/
{accession}.txtPlaintext · SEC filing body
{accession}.summaryAuto-generated TL;DR
watchlist/todayCurated · ticker-relevant only
wires/
global-newswire/today.feedRSS-ish stream
businesswire/today.feedRSS-ish stream
prompts/
premarket-briefPrompt · ready-made
earnings-recapPrompt · ready-made
filing-explainerPrompt · ready-made
Drop it into your agent today.
Free tier includes the full tool surface and 10k tool calls per month. One config line away.