# rfp agent hit rate limit mid answer scoring and failed to finish: how to fix it

An RFP agent that hits a rate limit mid scoring should checkpoint its progress and resume, not restart from zero. Add backoff with retry and per-answer checkpoints so a limit pause costs minutes, not the whole run. If your agent framework lacks checkpointing, batch smaller and save state between batches.

```text
rfp agent hit rate limit mid answer scoring and failed to finish
```

## When to use this skill
- An agent dies on rate limits mid scoring run
- Restarting redoes hours of already-scored work

## When not to use this skill
- The limit is on the RFP platform API, not the model (different skill)
- You need a higher quota (a billing conversation, not a bug)

## Tool compatibility
Agent frameworks with tool calling. Model APIs with rate limits.
Related area: model APIs and inference quotas.
Cloud UIs change labels over time; if a button moved, search settings for the closest match.

## Before you start
- Sign in to the tool and keep one tab open so sessions stay clean
- Copy the exact error text and note the time it happened
- Set aside 15 to 20 minutes; most of these resolve in a single pass

## Quick diagnosis
Make sure this is the right skill before you invest the 15 minutes.
- If you are dealing with an agent dies on rate limits mid scoring run, start at step 1 below.
- If instead the limit is on the RFP platform API, not the model (different skill), stop here; you need a different skill.
- If neither quite matches, compare your screen against the verbatim error block at the top; that block is exactly what this skill covers.

## Fix it in steps
### 1. Confirm the limit details

```
Read the rate limit error for the reset window and the quota that was hit.
```
Expected: a clear per-minute or per-day limit with a reset time.

### 2. Add checkpointing per answer

```
Change the agent to save each scored answer before moving to the next.
```
Expected: a restart resumes from the last saved answer.

### 3. Add backoff with retry

```
Wrap scoring calls in exponential backoff that waits out the limit window.
```
Expected: the agent pauses and resumes instead of dying.

### 4. Rerun to completion

```
Launch the run again and watch it survive the limit pause.
```
Expected: all answers scored with no rework.

## Confirm each step worked
Run through this checklist before moving on. If a check fails, redo that step instead of pushing ahead.
- Step 1 (confirm the limit details): Expected: a clear per-minute or per-day limit with a reset time.
- Step 2 (add checkpointing per answer): Expected: a restart resumes from the last saved answer.
- Step 3 (add backoff with retry): Expected: the agent pauses and resumes instead of dying.
- Step 4 (rerun to completion): Expected: all answers scored with no rework.

## Quick recap
1. Confirm the limit details
2. Add checkpointing per answer
3. Add backoff with retry
4. Rerun to completion

## Still stuck?
Gather three things: the exact error text, the time it happened, and what changed just before it broke.
Check the vendor status page for an active incident. Then open a support ticket with those three things;
it cuts the back-and-forth in half. For agent runs, attach the last 50 lines of the agent log too,
since the loop or timeout signature in the log usually names the cause.

## Variant phrasings
### agent rate limit mid run
Checkpoint plus backoff is the standard fix. If this matches what you saw, the steps above apply as written.

### rfp agent failed to finish scoring
Without checkpoints every retry starts over. If this matches what you saw, the steps above apply as written.

## Why it happens
Agents that score in one long loop hold no state, so a rate limit error kills the run and all progress. Checkpointing turns a fatal error into a pause, because the resume point survives the crash.

## Edge cases
- Backoff without jitter can synchronize retries into a new limit spike
- Very long runs should checkpoint to durable storage, not memory
- Track which quota you hit, per-minute and per-day need different pacing

## Provenance

Resolved from the public thread: https://vectle.com/posts/pst_ERRmVdUh17GiXZ8qd7voTA