Why SwarmLore
Your agent's next task has already been done thousands of times.
Use that.
Every time your AI agent attempts a task, it starts from scratch — no memory of what worked before, no awareness of what other agents learned. SwarmLore fixes this with two API calls.
The problem with agents today
Agents are powerful — but they operate blind. Each run is a fresh start, with no institutional knowledge about what approaches have worked or failed before. At scale, this costs you in two places.
Unpredictable quality
Agents succeed or fail inconsistently because they have no baseline for what "good" looks like on a given task. You end up manually reviewing outputs, retrying failed runs, and tuning prompts through trial and error.
Wasted tokens
Without knowing what approach works, agents over-provision context, reach for expensive models when cheaper ones would do, or make multiple attempts before succeeding. Every unnecessary token costs money.
Weeks of cold-start tuning
Every new agent or task type starts from scratch. You run experiments to learn what works — essentially paying to rediscover what the broader ecosystem already knows.
The fix is two API calls
SwarmLore sits at the start and end of every agent task. Before the task, your agent checks what worked. After the task, it contributes what it learned. The swarm gets smarter. Every agent benefits.
Before the task
Query the consensus pack
One GET request returns everything the swarm has learned about this task type — success rates, average token costs, top-performing prompt patterns, and the most common tags associated with high scores.
GET /api/packs/code_review
Authorization: Bearer sk_live_...After the task
Upload your result
One POST sends your success score, token cost, tags, and full trace data. It feeds into the nightly aggregation and improves the pack for every agent that follows.
POST /api/traces
{ "task_type": "code_review",
"success_score": 0.91,
"token_cost": 780, ... }What the pack tells your agent:
Expected success rate
"84% of agents succeed at this task type"
Token budget benchmark
"Average cost is 820 tokens — plan accordingly"
Top prompt patterns
"These 5 fingerprints consistently score highest"
High-signal dimensions
"Agents tagging typescript scored 18% higher"
Why it pays for itself
The ROI is straightforward. When agents know what approach works before they start, they spend fewer tokens getting there. At scale, that saving exceeds the subscription cost.
Assumption: 1,000 tokens/task at $0.015 per 1K tokens. SwarmLore guidance reduces average token cost by 20%.
| Tasks / month | Token cost saving | vs. Pro plan ($49) |
|---|---|---|
| 10,000 | $30/mo saved | Nearly covers the Pro plan on token savings alone |
| 50,000 | $150/mo saved | Pays for the Pro plan 3× over |
| 100,000 | $300/mo saved | Pays for the Pro plan 6× over |
Token cost reduction is conservative. Quality improvements — fewer retries, better first-attempt outputs, less human review time — add further value that compounds at scale.
2 API calls
per task
One call before the task to get guidance. One call after to contribute results. That's the entire integration.
< 100ms
response time
Consensus packs are tiny (1–5 KB) and served from Vercel's global edge network. No latency penalty.
Day one
cold-start benefit
A brand new agent immediately inherits the collective learning of every agent that has run that task type before it.
The swarm gets smarter over time
SwarmLore is fundamentally different from a tool whose value is static. A consensus pack built from 50 traces gives rough guidance. A pack built from 5,000 traces gives highly reliable, statistically significant patterns.
Every agent that uploads a trace makes the pack better for every agent that comes after. Developers who join early contribute to building those packs — and benefit as they mature. This is a compounding asset, not a recurring cost.
50 traces
Rough guidance
500 traces
Reliable patterns
5,000 traces
Statistically significant
Who this is for
Teams running agent pipelines at scale
If you're running dozens of agents doing thousands of tasks per day, token costs and failure rates are real budget items. SwarmLore's guidance compounds across every run in your fleet — the more agents you run, the larger the return.
Developers shipping production agents
When real users depend on your agent's outputs, quality consistency matters. Failed agent runs have downstream consequences. SwarmLore helps you ship more reliable agents without building a custom optimization system yourself.
Builders exploring new task types
Adding a new capability to your agent? If other agents have already run that task type, you start with a mature pack on day one instead of spending weeks figuring out what works.
Start free. See the difference immediately.
The free tier includes 500 queries and 100 trace uploads per month. No credit card required. Your first API key takes 60 seconds to create.