Short Answer
Of the 10 finance agent templates Anthropic released on May 5, 2026, four are useful for retail traders: the Earnings Reviewer, the Pitch Builder (for stock thesis writeups), the Financial Model Builder, and the Meeting Prep tool (rebrandable as a daily research brief). The other six target month-end close, KYC, and audit workflows that retail traders do not run. All four useful agents install as Claude Code plugins or work as cookbooks for Claude Managed Agents.
What Anthropic Actually Released
On May 5, 2026, Anthropic launched ten “agent templates” for financial services. Each one is a reference architecture that bundles three things:
| Layer | What It Does | Example |
|---|---|---|
| Skills | Instructions and domain knowledge for the task | ”Read a 10-K and extract guidance changes” |
| Connectors | Governed access to data sources | FactSet, Moody’s, Excel, internal filings |
| Subagents | Smaller Claude models for specific sub-tasks | Comparables selection, methodology checks |
The templates ship in three forms: as plugins inside Claude Cowork and Claude Code, and as cookbooks for Claude Managed Agents. The retail-friendly path is the Claude Code plugin, which you install with one command and run locally.
The full list of templates Anthropic shipped:
| Agent | Built For | Useful for Retail? |
|---|---|---|
| Pitch Builder | IB analysts drafting pitchbooks | Yes (as stock thesis writer) |
| Meeting Prep | Coverage bankers prepping for client meetings | Yes (as daily research brief) |
| Earnings Reviewer | Equity research updating models post-earnings | Yes |
| Financial Model Builder | Junior bankers building DCF and comps | Yes |
| General Ledger Reconciler | Corporate finance teams | No |
| Month-End Closer | Controllership | No |
| Financial Statement Auditor | Audit firms | No |
| KYC Screener | Compliance teams | No |
| Risk Memo Drafter | Credit risk | No |
| Investment Committee Packager | PE/VC IC prep | Edge case |
The Four Worth Installing
1. Earnings Reviewer
This is the one that justifies the whole release for retail traders. Point it at an earnings transcript URL or PDF, and it returns:
- Beats and misses against consensus
- Guidance changes (raised, lowered, reiterated)
- Specific quotes that shifted the narrative
- Suggested updates to your existing model assumptions
I tested it on NVDA’s most recent earnings call. The agent flagged the data center segment guidance bump, pulled the exact CFO quote, and listed three model assumptions to revisit (gross margin trajectory, capex, China revenue). Doing this manually takes 45 minutes. The agent took 90 seconds.
How to use it for trading: run it the morning after earnings, before the market opens. The summary tells you whether the price action overnight matches the fundamental shift or whether there is a gap.
2. Pitch Builder (Reframed as Stock Thesis)
The default pitchbook prompt is wrong for retail use. Reframe the prompt as “build a long-form stock thesis” and the agent does the same thing in a useful format:
- Business overview and segment breakdown
- Comparables table (auto-populated from your connector)
- Bull case, bear case, and consensus view
- DCF skeleton with editable assumptions
The output reads like a sell-side initiation report. For a retail trader running a concentrated portfolio of 5-10 stocks, having a thesis document on each name is useful for keeping conviction during drawdowns. We covered the broader pattern in the daily AI trading research routine.
3. Financial Model Builder
This one builds DCF and comparables models in Excel. With Anthropic’s new Claude add-ins for Microsoft 365 (announced the same day), the agent writes formulas directly into your workbook.
For a retail trader, the practical use case is sanity-checking your assumed valuation against the model’s. If you think AAPL is worth $250 based on a back-of-envelope DCF, run the agent with your inputs and see if the formal model agrees. Where it diverges is where your assumptions are weakest.
4. Meeting Prep (Reframed as Daily Brief)
The Meeting Prep agent reads your calendar, pulls relevant news on the company you are about to meet, and assembles a one-pager. Retail traders do not have client meetings, but the same workflow makes a great daily portfolio brief: feed it your current positions, and it pulls overnight news, analyst rating changes, and any 8-K filings on each name.
This replaces the manual scan you would otherwise do on a Bloomberg terminal or by tabbing through Yahoo Finance.
How to Install the Plugins in Claude Code
Anthropic published the templates to the financial-services repo on GitHub. Each agent has its own directory with a manifest, the skills, and the connector definitions.
To install one in Claude Code:
git clone https://github.com/anthropics/financial-services.git
cd financial-services/earnings-reviewer
claude /plugin install .After install, the agent shows up in Claude Code’s plugin list and you invoke it with /earnings-reviewer <transcript-url>. The first run prompts for any required connectors. For retail use, you can skip FactSet and point it at free sources (SEC EDGAR, company IR pages).
If you want to combine these with the broader Claude Code workflow, the Claude Code trading skills guide covers how skills compose with custom commands.
What These Are Not
A few things to set expectations:
- Not real-time signal generators. None of the ten agents place trades or read live market data. They work on filings, transcripts, and structured financial data.
- Not a replacement for a broker integration. If you want a bot that actually trades, see the Alpaca MCP tutorial or the Bybit MCP guide.
- Not free in production. Running the agents costs Claude API tokens. For retail use the cost is trivial (cents per earnings review), but a high-frequency setup will add up.
Comparison: Anthropic’s Agents vs DIY Skills
If you have already built custom Claude Code skills for trading research, the new templates may overlap with what you have. Here is the trade-off:
| Approach | Pros | Cons |
|---|---|---|
| Anthropic templates | Maintained, tested, integrated with M365 | Built for institutional shape, needs reframing |
| Custom skills | Tuned to your workflow, no reframing needed | You maintain them, no connectors out of the box |
| Hybrid | Use templates for heavy work, custom for orchestration | More moving parts |
Most retail traders should start with the templates and only build custom skills for the gaps. The Earnings Reviewer alone saves enough time that the install is worth it.
What Comes Next
Anthropic positioned this release as the first batch. Bloomberg’s coverage of the launch quoted the team saying more verticals are coming, including asset management and private markets. The interesting question for retail is whether a “buy-side research” template ships next, since that is closer to what an active retail trader actually does.
For now, install the four useful templates, skip the rest, and add the Earnings Reviewer to your weekly routine during earnings season. The other three are nice-to-have. The Earnings Reviewer is the one that pays for itself the first time you use it.
